Transnationalism and Integration of Turkish and Romanian Migrants in Western Europe

Dissertation zur Erlangung des Doktorgrades (Dr. phil.) des Fachbereichs Kultur- und Sozialwissenschaften der Universität Osnabrück

vorgelegt von

Steffen Pötzschke

Osnabrück, 2018

Contents

List of tables ...... 3 List of figures ...... 4 List of appendices ...... 4 Abbreviations and acronyms ...... 5 English abstract ...... 6 German abstract ...... 7 Acknowledgements ...... 8 1 Introduction ...... 9 2 Theoretical considerations ...... 17 2.1 Introductory remarks on methodological nationalism ...... 17 2.2 Assimilation and integration of migrants ...... 20 2.2.1 Differentiating between the political and the scientific understanding of terms ...... 20 2.2.2 The scientific understanding of assimilation and integration ...... 22 2.2.3 The relevance of context factors for the integration process ...... 34 2.3 Transnationalism in migration research ...... 35 2.3.1 Defining the phenomenon ...... 35 2.3.2 Migrant transnationalism in historical perspective ...... 39 2.3.3 Systematising transnationalism ...... 43 2.4 Transnationalism and integration – Mutually exclusive or coexisting? ...... 50 2.5 Summary and hypotheses ...... 61 3 Methods, research design, and data ...... 72 3.1 Methodological considerations ...... 72 3.2 Analytical strategy and employed techniques ...... 73 3.3 Survey design and fieldwork ...... 76 3.4 Measurement and operationalisation of key aspects ...... 79 3.4.1 General sociodemographic and migration-related information ...... 79 3.4.2 Aspects of transnationalism...... 82 3.4.3 Social integration ...... 93 3.4.4 Identificational integration ...... 95 4 Sociodemographic description ...... 96 4.1 Turkish and Romanian migrant population in Denmark, Germany, and Italy ...... 96 4.2 General migration patterns, age, gender, and migration motives ...... 97 4.3 Education, labour market participation, and economic situation ...... 105 4.4 Summary ...... 113 5 The transnationalism of Turkish and Romanian migrants ...... 115 5.1 Multidimensionality of transnationalism ...... 115 5.2 Transnational human capital ...... 121 5.3 Transnational background and networks ...... 123 5.4 Transnational practices ...... 131 5.4.1 Transnational communication ...... 131 5.4.2 Economic transnationalism ...... 134 5.4.3 Transnational mobility ...... 138 5.4.4 Political transnationalism ...... 149 5.4.5 Transnational consumption ...... 150 5.5 Determinants of transnational practices ...... 152 5.6 Summary ...... 170 6 The relationship between transnationalism and integration...... 174 6.1 Social integration ...... 174 6.2 Identificational integration ...... 187 6.2.1 Identificational integration in the country of residence ...... 190 6.2.2 Identificational integration in the town of residence ...... 198 6.3 Comparison and summary ...... 204 7 Conclusion ...... 210 Appendix ...... 222 Bibliography ...... 230

List of tables Table 1: Overview of hypotheses ...... 71 Table 2: Operationalisation of transnationalism ...... 83 Table 3: Basic demographic information ...... 103 Table 4: Migration motives ...... 104 Table 5: Highest educational title ...... 106 Table 6: Indicators used in the multiple correspondence analysis of transnationalism ...... 118 Table 7: Features of transnational human capital ...... 121 Table 8: Contacts abroad ...... 124 Table 9: Determinants of transnational practices ...... 156 Table 10: Ordinal logistic regression of social integration ...... 180 Table 11: Identification with country of residence, town of residence, and country of origin .... 187 Table 12: Linear regression of identification with the country of residence...... 191 Table 13: Linear regression of identification with the town of residence ...... 199 Table 14: Overview of effects on social and identificational integration ...... 204

List of figures Figure 1: Cumulative relative growth of samples by year of settlement in the CoR – Turkish migrants ...... 99 Figure 2: Cumulative relative growth of samples by year of settlement in the CoR – Romanian migrants ...... 101 Figure 3: Labour market participation ...... 109 Figure 4: Household financial situation ...... 111 Figure 5: Real estate ownership by household financial situation ...... 112 Figure 6: MCA plots of indicators within the dimensions and subdimensions of transnationalism ...... 117 Figure 7: Family, in-laws, and friends from the CoO living in the CoO ...... 125 Figure 8: Family, in-laws, and friends from the CoO living in a third country ...... 125 Figure 9: Family, in-laws, and friends from a third country not living in the CoR ...... 125 Figure 10: Geographic scope of transnational networks ...... 128 Figure 11: Use of different means of transnational communication ...... 132 Figure 12: Respondents sending money abroad ...... 134 Figure 13: At least one trip abroad during the last 24 months ...... 140 Figure 14: Three or more trips abroad during the last 24 months ...... 140 Figure 15: Geographic scope of transnational mobility ...... 145 Figure 16: Features of transnational consumption ...... 151 Figure 17: Family, in-laws, and friends from the CoR living in the CoR ...... 175 Figure 18: Family, in-laws, and friends from the CoO living in the CoR ...... 175 Figure 19: Family, in-laws, and friends from a third country living in the CoR ...... 175

List of appendices Appendix 1: Classification of educational titles ...... 223 Appendix 2: Official population statistics on Turkish and Romanian migrants in Denmark, Germany, and Italy ...... 224 Appendix 3: Bivariate relation between migration motives and year of settlement ...... 224 Appendix 4: Transnational networks – most often mentioned countries ...... 225 Appendix 5: Transnational mobility – most often mentioned countries ...... 225 Appendix 6: Independent variables used in regression models ...... 226 Appendix 7: Regression of social integration and identificational integration using transnationalism indices ...... 228

Abbreviations and acronyms

EUCROSS The Europeanisation of Everyday Life: Cross-Border Practices and Transnational Identities among EU and Third Country Citizens (research project)

CoO/ CoOs Country of origin/ countries of origin

CoR/ CoRs Country of residence/ countries of residence

ToR/ ToRs Town of residence/ towns of residence

TU in DK Turkish migrants in Denmark

TU in DE Turkish migrants in Germany

TU in IT Turkish migrants in Italy

RO in DK Romanian migrants in Denmark

RO in DE Romanian migrants in Germany

RO in IT Romanian migrants in Italy

CIEP The Comparative Immigrant Entrepreneurship Project

EU European Union

GDP Gross domestic product

ISSP International Social Survey Program

SNS Social networking sites

SOEP (German) Socio-Economic Panel

VoIP Voice over Internet Protocol

English abstract

In today’s world, cultural, social, and economic relationships increasingly transcend national borders. This coincides with high levels of migration and physical mobility across the globe. Therefore, this study investigates the degree that migrants participate in such transnational phenomena and how this participation relates to their integration in the country of residence. To address these questions, the cross-border relationships and practices of Turkish and Romanian migrants in Denmark, Germany, and Italy are examined. The study analyses the degree to which migrants are transnational, by which factors migrants’ transnationalism is determined, and whether transnationalism impacts the social and identificational integration of migrants in their country of residence. With respect to identification, not only the national but also the local level is taken into focus.

The use of a unique quantitative dataset allows the study to address the often-voiced need for further quantitative findings on migrants’ transnationalism in Europe and particularly the lack of cross-national studies in this field. Additionally, the analysis goes beyond migration research’s typical focus on transnational connections between migrants’ country of origin and country of residence.

The results show that the level of migrants’ transnationalism strongly differs between the measured dimensions of this phenomenon. Moreover, transnational connections and practices of all samples go beyond the dichotomy of origin and residence country, stressing the need to incorporate a broader perspective in migration research. The analysis does not find many effects of transnationalism on social and identificational integration. The few observable effects are in part of a positive and a negative nature. Furthermore, these effects differ across the analysed dimensions of integration. Based on these results, the study highlights that transnationalism and integration are not mutually exclusive.

German abstract

Kulturelle, soziale und wirtschaftliche Beziehungen überwinden heute zunehmend nationale Grenzen. Gleichzeitig befinden sich Migration und Mobilität weltweit auf einem hohen Niveau. Die vorliegende Studie untersucht, inwiefern Migranten Teil solcher transnationaler Phänomene sind und wie sich diese Teilhabe auf ihre Integration in das Aufenthaltsland auswirkt. Um diese Fragen zu beantworten, werden die grenzüberschreitenden Beziehungen und Praktiken türkischer und rumänischer Migranten in Dänemark, Deutschland und Italien untersucht. Die Studie analysiert das Transnationalitätsniveau der Migranten, von welchen Faktoren ihr Transnationalismus beeinflusst wird und ob Transnationalismus seinerseits einen Einfluss auf die soziale und identifikative Integration in das Aufenthaltsland hat. Bei der Untersuchung der Identifikation wird nicht nur die nationale, sondern auch die lokale Ebene in den Blick genommen.

Durch die Nutzung eines einzigartigen quantitativen Datensatzes kann die Studie den oft betonten Bedarf an zusätzlichen quantitativen Forschungsergebnissen zum Transnationalismus in Europa adressieren. Insbesondere gilt dies auch für den Mangel an kulturell vergleichenden Studien in diesem Bereich. Außerdem geht die Analyse über den üblichen Fokus der Migrationsforschung auf transnationale Verbindungen zwischen dem Herkunfts- und Aufenthaltsland hinaus.

Die Ergebnisse zeigen, dass sich der Umfang in dem Migranten transnational sind, zwischen den einzelnen Dimensionen dieses Phänomens stark unterscheidet. Es wird zudem deutlich, dass transnationale Beziehungen und Praktiken in allen Befragtengruppen über die Dichotomie von Herkunfts- und Aufenthaltsland hinausgehen. Dies unterstreicht, dass die Migrationsforschung in dieser Hinsicht ihre Perspektive erweitern sollte. In der Analyse werden nur wenige Beziehungen zwischen Transnationalismus und sozialer sowie identifikativer Integration offenbar. Die wenigen gefundenen Effekte sind sowohl positiver als auch negativer Natur, unterscheiden sich aber zudem zwischen den Integrationsdimensionen. Vor dem Hintergrund dieser Ergebnisse unterstreicht die Studie, dass sich Transnationalismus und Integration nicht gegenseitig ausschließen.

Acknowledgements

First and foremost, I would like to thank my advisers Prof. Dr. Andreas Pott and Prof. Dr. Michael Braun for their support and invaluable advice during the writing of this thesis. Without their guidance, I would have been lost more than once along the way. Of course, I also want to thank Michael Braun for giving me the opportunity to participate in the EUCROSS project, in the first place. Without this great project, this thesis would not have been possible. I am deeply grateful to my colleagues in our international research consortium. Furthermore, I would like to thank the European Commission for funding the project as part of the 7th Framework Programme.

Throughout the years, I benefitted from the stimulating work environment and academic freedom provided by the GESIS - Leibniz Institute for the Social Sciences. I am very grateful to all my colleagues here for their support and encouragement. This goes without saying, especially for all members of the Cross-Cultural Surveys and the GESIS Panel team. I am extremely grateful to Dr. Dorothée Behr and Wanda Otto for their valuable input and critical feedback to earlier versions of this thesis. Finally, I would like to express my gratitude to my whole family for their encouragement and help throughout the years. I thank my parents, who have assisted me in many ways. Without Carolina’s practical and emotional support I would certainly not have been able to realise this work.

1 Introduction

Migration has been a key element of human history since its beginning. Homo sapiens populated the world as Homo migrans, as the historian Klaus Bade (2003, ix) put it. According to estimations of the International Organization for Migration (2016, 5), 244 million people were living outside their country of birth in 2015. The reasons for migration are as diverse as the people who engage in it. Some might move to study abroad or to start a career, others may do so to live with loved ones, and far too many are compelled to migrate to escape poverty, environmental disasters or violent conflicts. While none of these reasons, or the many others imaginable, are new per se, they currently coincide with a situation of increased globalisation. This growing connectivity of geographically distant places, people, and events is visible in many areas of daily life. People and cultures are increasingly connected through electronic means of communication, and information can cross distances in seconds which would have taken weeks a century ago. The same holds true for the consumption of goods. Thanks to the internet, it is easier than ever to buy commodities in other countries and have them sent to almost any part of the world. Of course, these opportunities are highly dependent on the existence of a reliable infrastructure and not accessible to the same degree in all countries. While both limitations also affect long-distance travel, it is, nevertheless, available to larger parts of the world’s population, more affordable, and significantly faster than in previous centuries. In this situation, the question arises of whether and how these possibilities to traverse borders physically and non-physically at comparably low cost might influence the daily life of migrants. More precisely: To which degree do migrants use the existing possibilities to stay connected to their country of origin (CoO) or to friends and family in third countries? What practices do they engage in? What are the possible determinants of such engagement? And finally, how do such practices and connections relate to their lives in their country of residence (CoR)?

Using data collected on Turkish and Romanian migrants in Denmark, Germany and Italy, this study will investigate these issues by concentrating on three key topics: First, the degree to which migrants are transnational will be investigated. This involves examining whether migrants are part of social networks spanning international borders, whether they engage in cross-border practices, and to which extent they possess skills and knowledge that would allow them to do so. A comparative point of view will be taken in this regard as the data stem from migrants from two different countries of origin in three countries of residence. Second, it is important to understand

Introduction which factors foster or impede migrants’ engagement in transnational activities. Finally, the potential influence of transnationalism on migrants’ integration into their CoR will be explored. The work aims to provide empirical findings on transnationalism in general, and on its relationship with integration in particular. However, it is not the goal to develop a comprehensive theory of transnationalism or a model that could explain integration as such.

During the last 30 years, there has been a major shift in migration research. While earlier studies were mainly concerned with the circumstances and consequences of immigration, both with regard to migrants and the receiving countries in a broader sense, scholarly interest has since started to take immigrants’ continuous cross-border ties into view. Enduring connections between migrants and those they leave behind are not without historical precedent and have been the subject of prior research (see Morawska 2003; Foner 2007). However, driven mostly by technological improvements, higher accessibility, and cost reductions in the sectors of communication and transport, the nature of social connections kept across borders and the practices arising in this changed context differ from earlier periods. The same holds true for the number of migrants who engage in these practices (Faist, Fauser, and Reisenauer 2013, 50-51). These cross-border connections and activities are summarised under the term transnationalism and have been studied by numerous scholars. Important foci are: the political participation of emigrants and their inclusion in political structures of the CoO (e.g., Bauböck 2003; Itzigsohn and Villacrés 2008; Østergaard-Nielsen 2003); migrants’ individual and collective economic activities (e.g., Portes 2009; Escobar 2010); and migrants’ social interactions with people in their places of origin while being abroad and during visits (e.g., Goldring 1999). Groundbreaking research on transnationalism started in the early 1990s. It was, at that time, of a mainly qualitative nature and mostly focussed on Latin-American migrants in the United States. The resulting studies are without any doubt of immense value. They provide in-depth information on the role that transnationalism plays in migrants’ lives and the circumstances enabling and promoting it in the specific cases under observation. However, quantitative findings, which allow for a more general evaluation of the determinants and effects of transnationalism, were in short supply until more recently, as various authors hitherto pointed out (e.g., Snel, Engbersen, and Leerkes 2006, 303; Kivisto and Faist 2009, 149; Bilgili 2014a, 286). Such results are, furthermore, still scarce on a cross-national scale.

Introduction

Besides the general need for further research on the nature of transnationalism, the question arises of how transnationalism relates to the participation and inclusion of migrants in the social systems of their CoR. It is the latter issue, i.e., whether and how immigrants become part of the receiving country and participate in its vital social systems, which has been central to migration research since the first half of the 20th century. A variety of different terms are used to describe these processes. Most prominent amongst them are assimilation and integration. Concepts of assimilation were first popularised by scholars of the Chicago School of Sociology (e.g., Park and Burgess 1921) and reached considerable influence amongst migration researchers in the United States and beyond. Common to these early theories was the assumption that immigrants would gradually focus their lives entirely in their country of residence and cut all significant ties to their country and culture of origin. After a decline in the wake of scholarly critique starting in the late 1960s, there has been renewed interest in assimilation theories in the last 25 years. Proposed modifications of earlier concepts include, for example, the segmented assimilation theory put forward by Portes and Zhou (1993) and the new assimilation theory by Alba and Nee (2003). Both the classical and revised assimilation concepts continue to have a significant influence on migration research in European countries. However, European scholars usually employ the term integration instead of assimilation to describe societal participation and inclusion of immigrants in the receiving country. A particular influential integration theory is the one proposed by Esser (1980; 2000). In his work, this author directly builds on the above-described concepts of assimilation. There are, however, important differences in terminology. Esser generally speaks of integration, and while the term assimilation is important to his theory, too, it constitutes only one of four paths of integration that are theoretically possible according to his model.

In addition to addressing the above-stated research questions, this study aims to make four important contributions to transnationalism research. The first consists in the use of a quantitative cross-national dataset. While several quantitative studies investigating transnationalism have been published in recent years, their observations pertain only to a small number of countries, most notably to the United States and the Netherlands. Furthermore, most of these studies concentrate on one specific CoR at a time. Consequently, cross-national comparisons of the nature and determinants of transnationalism can only be made using the data of different studies. However, this is problematic as available studies collected data with different instruments using varying operationalisations of transnationalism. Indeed, Schunck (2014, 292) points out that

Introduction studies employing a genuine cross-national research design are needed to overcome this limitation. Through its use of quantitative data collected in a single study on two migrant populations in three European countries, this work will help to narrow this research gap.

Second, in addition to describing the phenomenon of transnationalism in general, more quantitative research is needed so that scholars could gain a better understanding of the relationship between transnationalism and integration. Existing assessments of this relationship vary in their conclusions. From the perspective of assimilation and integration theory, these phenomena (i.e., integration/assimilation and transnationalism) are perceived to be mostly incompatible. Esser, for instance, describes the exclusive concentration of immigrants on the CoR context, i.e., assimilation, as the only empirically relevant type of successful integration. Multiple integration is seen by him as a mostly theoretical category and could, if at all, be achieved only by a small minority of the migrant population, such as children of diplomats, academics and artists (Esser 2001, 21; Esser 2003, 7-9). Consequently, according to this theory, the simultaneous integration of immigrants in the CoR and in transnational networks, which could be described as a form of multiple integration, is unlikely and not empirically relevant. Indeed, early conceptualisations of transnationalism likewise depicted this phenomenon as an alternative to assimilation and implied that the two were diametrically opposed (e.g., Glick Schiller, Basch, and Blanc-Szaton 1992b, 16; Portes, Guarnizo, and Landolt 1999, 228-29). However, in recent years the growing consensus among scholars in this field is that transnationalism and integration are not mutually exclusive, but might even be interrelated (Morawska 2002, 161-62; Levitt 2002, 192; Faist, Fauser, and Reisenauer 2013, 100). As mentioned before, quantitative findings to support (or question) this assessment are small in number. Some evidence is provided by Guarnizo, Portes, and Haller (2003, 1238) who found that political transnationalism is not associated with marginalisation in the CoR. Likewise, Carling and Hoelscher (2013) showed that structural integration in the receiving context is actually positively related to transnational economic engagement. As discussed in the second chapter, most other studies come to similar results and do not find an overall negative relation between integration and transnationalism. However, this does not mean that such evidence would be completely inexistent. For instance, Ley (2013) discovered that naturalised migrants who identified with their new country were less likely to engage in transnational activities. The author himself stresses that this finding differs from other studies (Ley 2013, 935) and points to the different origin countries of the migrants and

Introduction general differences of study design as possible explanations. The data used in this work will help to avoid these problems, as individuals from the same CoOs will be studied in different CoRs. By including CoO and CoR as control variables, the analysis will also be able to investigate whether they are relevant determinants of transnationalism and integration. While the discussion often concentrates on positive or negative associations, it is also possible that aspects of transnationalism and integration may lack any significant relation with each other. In fact, quantitative studies often found that many of their transnationalism indicators showed no effect on integration (e.g., Snel, Engbersen, and Leerkes 2006, 304; Schunck 2014, 288).

The third important contribution of this work concerns the geographic focus employed in research on migrant transnationalism. While some authors indicate that migrant transnationalism is not necessarily limited to connections between people or institutions in only two countries (e.g., Faist, Fauser, and Reisenauer 2013, 54), current research typically concentrates on connections forged and sustained between the origin and residence country of migrants. It is certainly true that transnational networks and activities of individuals are most likely to link the country they originated from with the country they reside in; however, it is not clear why migrant transnationalism should be limited to these two countries only. Indeed a geographically more diverse nature of the phenomenon can be expected for several reasons: First, migrants are likely to know other migrants, such as individuals from their CoO who might live in third a country or immigrants of other nationalities who might incorporate them in their own networks. Second, migrants may have previously lived in another country before settling in their current CoR. Finally, and most importantly, recent research has shown that transnationalism is not limited to migrants, but that parts of non-migrant populations in EU countries also engage in transnational activities and form part of transnational networks, even though this is clearly not a majority phenomenon (Mau 2010; Kuhn 2015). Since the sedentary populations surveyed for these studies lived in their CoO, their transnationalism cannot be described within a CoR-CoO framework. There is no reason to assume that migrants are any different from non-migrants in this respect. Consequently, this study will employ a broader geographic focus than most research on migrant transnationalism by also investigating networks and activities that are not limited to the dichotomy between CoO and CoR.

The fourth contribution concerns the type and variety of transnationalism indicators used in this analysis. Three aspects can be differentiated in this regard. First, research often shows that

Introduction different features of transnationalism are not determined in the same way by comparable sets of variables. This means a predictor might influence a first aspect of transnationalism positively and a second negatively, whilst lacking any significant association with a third aspect. Furthermore, it is regularly found that indicators of transnationalism themselves differ in their relation with aspects of integration very much in the same way (e.g., Snel, Engbersen, and Leerkes 2006; Schunck 2014). Consequently, it is advisable to include a variety of indicators to account for these differences when transnationalism serves as dependent or independent variable. A second related issue refers to the challenge of cross-national comparisons, in a situation in which such comparisons are only possible by pooling the results of different studies. A literature review quickly shows that published studies investigate a variety of transnationalism aspects and rarely try to answer the exact same research question. Consequently, cross-national comparisons that build on their findings are limited to the overlap between these publications. This overlap, however, is likely to become smaller, the more studies are taken into consideration. Nevertheless, results from different countries are certainly available for some features of transnationalism, such as money transfers (e.g., Itzigsohn and Giorguli Saucedo 2002; Snel, Engbersen, and Leerkes 2006; Carling and Hoelscher 2013) and visits to the CoO (e.g., Portes, Guarnizo, and Haller 2002; O'Flaherty, Skrbis, and Tranter 2007; Schunck 2014). However, in these cases cross- national comparisons are often complicated by diverging operationalisations. Finally, while research on sedentary populations shows that their engagement in transnational activities is strongly influenced by transnational background, networks, and human capital (e.g., language knowledge) (Mau 2010; Kuhn 2015), these aspects are thus far not systematically included in the study of migrant transnationalism. This study addresses this shortcoming by using a multidimensional operationalisation of transnationalism (see below). Overall, this study compares favourably to the literature by including a wide array of transnationalism indicators to account for the heterogeneous character of transnationalism itself, facilitating cross-national comparisons, and enabling the investigation of the interrelationship between different dimensions of the phenomenon.

The data used in this dissertation stem from the research project “The Europeanisation of Everyday Life: Cross-Border Practices and Transnational Identities among EU and Third Country Citizens” (EUCROSS 2016), which was funded by the 7th Framework Programme of the

Introduction

European Commission. The analysis will focus on survey data collected from Turkish and Romanian migrants in Denmark, Germany and Italy. The same methodology and survey instruments were used in all countries, securing a very high level of comparability in this project.

The analysis will use a multidimensional understanding of transnationalism that combines an operationalisation proposed by Theresa Kuhn (2011, 814) with systematisations of transnational activities more broadly used in migration research. The dimensions considered are transnational human capital, transnational background and networks, and transnational practices. For the last category, further five subdimensions will be used, which are communication, economic activities, mobility, political participation and consumption. As mentioned, analysis on migrant transnationalism hitherto concentrated mostly on transnational practices and took the other two dimensions not systematically into consideration. It is, however, promising to widen this perspective as it can be assumed that an individual’s networks and human capital influence the probability to engage in transnational activities. Moreover, it is the intention to use a multidimensional operationalisation that could also be applied to the analysis of transnationalism amongst sedentary populations. In doing so, this work aims to facilitate future research investigating transnationalism irrespective of the mobility status of individuals.

In the analysis, integration is understood as the process in which immigrants engage with institutions, social structures, norms, and the population of the country of residences in order to find and “‘secure a place’ for themselves” (Penninx and Garcés-Mascareñas 2016, 11) in their living environment. Like transnationalism, integration is considered a multidimensional phenomenon. Following Esser (1980; 2000), the subprocesses of cultural, social,1 structural, and identificational integration are distinguished. If quantitative studies on the relationship between integration and transnationalism are reviewed with this categorisation in mind, it becomes obvious that the aspects that are most often included in the analyses fall into the realms of structural and cultural integration. On the contrary, features of social and identificational integration are included far less often in the analyses and discussions (see chapter 2.4). To address first and foremost those areas where additional information is needed, this study will concentrate on these two dimensions. Nevertheless, indicators of cultural and structural integration will be included as independent variables, since they are likely to influence

1 Esser calls this dimension ‘interaction’ and uses social integration as an umbrella term in a somewhat different way (see chapter 2.2).

Introduction integration in the observed dimensions. Furthermore, research indicates that the children of immigrants identify more strongly with their town than with their country of residence (Crul and Schneider 2010, 1262). The data suggest that the same is true for immigrants themselves. Consequently, the analysis will include both levels of identification which will allow investigating whether they differ in their relationship with transnationalism.

Following this introduction, the next chapter will lay out the theoretical foundation for the subsequent empirical analysis. First, some general remarks regarding the role of the nation state concept in research are given. This is followed by a discussion of assimilation and integration concepts, and an overview of transnationalism research. The main aim of these subchapters is to provide and contextualise the key definitions of this work. The chapter will then discuss previous findings on the relationship between transnationalism and integration, and present the guiding hypotheses of the analysis. The third chapter will provide details on the data, relevant survey design issues, and the methods and operationalisations used. The fourth chapter will give information on the sociodemographic composition and migration history of the samples, which will help to contextualise the remainder of the discussion. Chapters five and six comprise the main analysis of this study, with the fifth chapter focusing on transnationalism. It does so by, first, examining the relationship between the indicators within the various dimensions and subdimensions of transnationalism. This serves to empirically confirm the validity of these categories. Next, a detailed examination of migrants’ involvement in the different aspects of transnationalism is presented. In a last step, the chapter analyses the determinants of transnational engagement. The sixth chapter turns the attention of the reader to the question of how transnationalism might influence social integration and identification integration on a local and national level. Finally, the thesis closes with a summary chapter in which the key findings are highlighted.

2 Theoretical considerations

This chapter provides the background for the empirical analysis in this work. After some general thoughts on methodological nationalism, it introduces the two central concepts, transnationalism and integration, and examines previous findings regarding their relation.

2.1 Introductory remarks on methodological nationalism The question of whether and how continuous cross-border relations and activities of migrants influence their integration in the country of residence (CoR) is central to the present study. Therefore, it is appropriate to address a key discussion related to this topic before defining the two principal theories themselves. This debate concerns the question of how the idea of nation states has influenced migration research, and to which degree territorial states should be present in the analysis of migration at a time which is characterised by growing globalisation and transnationalisation.

The discussion of territorial states’ role in research is, unsurprisingly, not limited to the study of human mobility. In a highly cited article, John Agnew argued in 1994 that political theory was at the time captured in what he called a ‘territorial trap’. Andreas Wimmer and Nina Glick Schiller (2002) observed similar tendencies with respect to social sciences in general and migration research in particular. In both texts, the authors argue that researchers in these disciplines tended to largely ignore the genesis of the nation state as a concept and reality which is not only socially constructed but has also just emerged during the last two and a half centuries from processes which can be described as conflict-full, at best. Instead of reflecting this reality researchers would mostly view “the territorial state as existing prior to and as a container of society” (Agnew 1994, 77), attributing a natural character to the nation state (Wimmer and Glick Schiller 2002, 304). Hence, researchers would base their elaborations on a concept of culturally and, in many cases, ethnically homogenous society, with all its aspects contained within the borders of the respective territorial state. These tendencies are the core aspects of what Wimmer and Glick Schiller call methodological nationalism and describe as “the assumption that the nation/state/society is the natural social and political form of the modern world." (Wimmer and Glick Schiller 2002, 302).2

2 The term itself is, however, not a creation of Wimmer and Glick Schiller but – as they point out (2003, 576n2) – borrowed from Herminio Martins. Martins, in turn, criticised already in 1974 in a short passage he wrote on the topic “the neglect of trans-national [sic] phenomena as such” in social theory (Martins 1974, 277).

Theoretical considerations

Building on this definition, Faist et al. (2013, 138) summarise the problem as an undue focus on territorial states defined in national terms, which influences all aspects of the research process, from planning and study design to strategies of analyses and the interpretation of results. As Faist and co-authors further explain, this focus has the consequence that migration research usually concentrates on migration regimes and migrant integration in specific nation states, or on the comparison of geographically constrained spaces, without taking into account border-crossing social relations. The result is a limitation of the analyses to CoR related issues and an understanding of social practices and formations in which those are confined to specific geographic areas, separated by international borders. This tendency is evident in many studies relying on ‘classical’ assimilation and integration concepts, such as those presented in the next subchapter (2.2). In contrast, Wimmer, Glick Schiller, Agnew and others call for a broader research perspective which does not place emphasis on states as the only framework in which social interactions could unfold and should be studied. However, this does not mean that the importance of states – or the immigration context in a more general sense – would be dismissed altogether by those authors and this school of thought.

The argument in this regard is twofold: while empirical evidence, on the one hand, shows that neither social life nor economic processes are nowadays cut off by international frontiers, on the other hand, it is also true that states and their institutions remain significant for migrants. Their polities control access to – and in some cases exit from – sovereign territories and regulate important fields such as the labour market, educational systems, and healthcare schemes. The continuing relevance of territorial states is underlined by Pries and Seeliger who argue:

“Despite numerous signs of the erosion of the nation-state-based paradigm, we can still identify many speciﬁc references to concrete localities in the case of social structures, as well as in daily phenomena such as locally bound families, working arrangements, languages, social networks and national or regional identities, all of which contribute to providing social order through subjective structures of belonging.” (Pries and Seeliger 2012, 223)

At the same time the continued relevance of territorial states does not imply the existence of homogenous national cultures or societies nor does it suggest that cultural aspects or social relations are limited only to the territories which are governed by them. Steffen Mau (2010) for example, showed that individuals who were born in a particular country and never lived in another one could just as well be part of transnational social networks as they belonged to

Theoretical considerations different social groups (e.g., family and friendship networks) within the CoR. It is in this sense that Amelina and collaborators argue that states should be seen as “one of several possible social contexts within which to empirically analyze social relations, institutions, cultures, spaces, ethnicities and histories.” (Amelina et al. 2012, 2). To this end, various authors propose a transnational perspective which allows the consideration of a plurality of social contexts. These approaches will be discussed in the third subchapter (2.3).

Hence, the critique of methodological nationalism does not negate the relevance of territorial states and social formations within the countries of residence. However, it is deemed essential that researchers take into consideration that the spatial, administrative, and political categories they use are socially constructed. Therefore, it cannot be assumed that lines drawn on a map to separate such spaces from each other would naturally also cut through all social processes and constrain them to either side of the border. While using a transnational focus to counteract tendencies of methodological nationalism it has, of course, to be considered that concepts of space which are applied in this context (e.g., transnational social spaces, see subchapter 2.3.1) are social constructs too (Glasze and Pott 2014).

The question is which consequences could be drawn from the above for the present analyses and the arguments to be developed in this text. First, it has to be mentioned that the data used in this study stem from a research project which surveyed different populations in various countries. In these countries, members of the two target groups (Romanian and Turkish migrants) were identified by their country of birth and nationality during the sampling process.3 Consequently, country of residence and country of origin (CoO) are categories which, to some extent, structure the data. In accordance with the arguments above, their inclusion in the analysis is considered appropriate since the migrants originated either from an EU member state (Romania) or a third country (Turkey) which might mean that they have been subject to policies that differed both within their CoR and across countries. Furthermore, the inclusion of these categories allows relating the results directly to earlier findings by other scholars. However, this does not mean that the artificial nature of these categories is irrelevant. On the contrary, the latter underlines the need to include them as independent variables in the analyses. By doing so, it will be possible to ascertain whether they constitute relevant factors that significantly influence individual transnationalism and integration. Finally, this work is driven by the basic assumption that social

3 Detailed methodological information is provided in chapter 3.3.

Theoretical considerations relations and activities are not cut off by international borders. In this sense, the question of how border-crossing relations and activities are related to integration presupposes that social processes constituting migrants lives are not taking place in mutually exclusive national containers.

The next two subchapters will discuss the role of integration and transnationalism as key concepts in this analysis. The above already indicated that transnationalism research is, in comparison to integration or assimilation research, a relatively new field. Furthermore, its principal argument has been largely formulated as a critique of the underlying container logic of ‘classical’ integration and assimilation concepts. The present discussion will, therefore, first focus on the concepts of assimilation and integration at a general level, and on the understanding of integration used in the subsequent analysis. Next, transnationalism will be discussed in the same manner. Finally, an additional subchapter will consider empirical findings that relate both phenomena to each other. The chapter will conclude with a short summary and hypotheses statement, building on prior arguments and results from existing literature.

2.2 Assimilation and integration of migrants

2.2.1 Differentiating between the political and the scientific understanding of terms When trying to define integration, the first issue is the need to differentiate it from assimilation, as both terms are used frequently and sometimes even synonymous. Unfortunately, there is no consensus regarding the conceptualisation and understanding of either term in social research on an international level. In fact, Schunck asserts that there are up to 30 different expressions which are used in the research literature to describe this phenomenon (2014, 10). This plurality of concepts and meanings can be seen as a symptom of what Leo Lucassen once depicted as migration research’s “Babylonic confusion with regard to the key terms” (2006, 17). The fact that such expressions as integration and assimilation are employed not only in scientific discussions but also in other contexts contributes massively to this confusion. Therefore, researchers have repeatedly pointed out the need to distinguish between the use of these concepts as analytic tools within scientific research, and as political concepts or programmes (Lucassen 2006, 17-18; Erdal and Oeppen 2013, 870; Schunck 2014, 11).

While in the former context these terms are used to describe empirical observations, they are used in the latter to articulate political objectives and normative expectations. Actors who employ the idea of immigrant integration in public and political debates often also imply the existence of a

Theoretical considerations specific and clearly defined national culture, which immigrants should use to orientate and adapt themselves to, much in the sense of the above-described container concept. An example is the dubious notion of “Leitkultur” (guiding national culture, see Habermas 2010) which is frequently mentioned in the context of immigrant integration by politicians and commentators in Germany. Integration and assimilation as political concepts, furthermore, often imply that there is a specific point or state that immigrants could and should reach in order to be considered well integrated. This regularly results in “the politically loaded idea of integration as an identifiable ‘endpoint’ that social policy can implement” (Erdal and Oeppen 2013, 870). The link to social policies in this quote is highly relevant. After all, the non-scientific meaning of principal terms in this field is also heavily influenced by political measures on different institutional levels, for example by so-called integration policies. Consequently, the meaning of these expressions and political concepts is potentially subject to substantial changes over time. Finally, some of these terms are also connected to specific policies in the past, which might burden them in the present. This is especially true regarding assimilation, which, in the Western-European context, is connected to policies of linguistic and cultural homogenisation (Brubaker 2001, 533; Kastoryano 2002, 43; Schunck 2014, 10).

In contrast to the understanding of these two terms in public discourse, both integration (see, e.g., Bade and Bommes 2004; Erdal and Oeppen 2013; Alba and Foner 2015b) and assimilation (see, e.g., Esser 2001; Alba and Nee 2003; Schunck 2014) are primarily perceived and conceptualised as continuing processes in contemporary academic contributions. Hence, while in the public discourse fixed – albeit often unspecific – criteria are formulated, and the question is asked to what degree migrants have achieved those static goals, the academic discussion concentrates more strongly on the observation of the dynamic processes. This does, of course, not preclude the task of assessing integration or assimilation levels at a specific point in time. However, academic rigour demands for such analysis and discussions to be conducted in a non-normative manner. Hence, terms such as integration and assimilation should be used scientifically to describe specific empirically observable phenomena and not desired outcomes. As various authors argue, this absence of normative assessments is, however, not a natural given in science either, but must be achieved through self-reflection and the readiness to re-evaluate established theories and schools of thought (e.g., Lucassen 2006, 16-18; Schunck 2014, 10-11).

Theoretical considerations

2.2.2 The scientific understanding of assimilation and integration In the context of migration research, the word assimilation is mostly used within academic circles in the United States, where it was popularised by members of the Chicago School of Sociology (Lucassen 1997, 8; Faist 2001, 31; Alba and Nee 2003, 63). Various scholars who are considered part of this school started to systematically study the settlement and incorporation processes of immigrants in the United States during the first quarter of the 20th century. Based on their observations they formulated theories of immigrant assimilation which were continuously refined within this circle until the early 1960’s. In the book “Introduction to the science of sociology”, Robert E. Parks and Ernest W. Burgess describe assimilation as a concept which becomes relevant in the context of immigration and proceed by defining it as follows:

“Assimilation is a process of interpenetration and fusion in which persons and groups acquire the memories, sentiments, and attitudes of other persons or groups, and, by sharing their experience and history, are incorporated with them in a common cultural life.” (Park and Burgess 1921, 735)

This early definition already indicates the central position which cultural incorporation (i.e., acculturation) takes in assimilation concepts developed to describe the immigration situation in the United States. A frequently mentioned limitation of these early concepts is that the acquisition of new attitudes, skills, and customs, i.e., the adaption processes, were almost exclusively conceived to be experiences of the newcomers. Consequently, these theories mostly disregarded the possibility that the social groups and settings in which this incorporation takes place might be simultaneously altered by this process as well (Alba and Nee 2003, 64). Furthermore, these early approaches considered assimilation as a natural and inevitable occurrence, especially in respect to immigration processes to the United States at the time (Park 1921, 758; Alba and Nee 2003, 65). Hence, they asserted that immigrants and their descendants would ultimately become part of a rather homogenous majority group. Another defining aspect is that assimilation was usually conceived as a linear sequence of subprocesses which followed a particular order. This general line of thought can still be seen in the work of Milton M. Gordon (1964, 70-71) who differentiated various aspects of assimilation, namely cultural, structural, marital, identificational, attitude receptional, behaviour receptional, and civic assimilation. Somewhat diverging from earlier approaches Gordon mentions the general possibility of partial assimilation, meaning that it could occur in some of the dimensions without being present in others (e.g., cultural and civic

Theoretical considerations assimilation only). However, he also stresses that if structural and cultural assimilation are given “all of the other types of assimilation will naturally follow.” (1964, 81).

The appeal of assimilation theory varied over time. After initial enthusiasm, the usage of the concept declined starting from the mid-1960s onwards. This development was partly due to growing evidence of persisting ethnic affiliations in some migrant groups and was partly rooted in a growing rejection of the political assimilation concept amongst academics (Brubaker 2001; Kivisto and Faist 2009). However, a revival of assimilation theory can be observed since the last decade of the 20th century. One of the most prominent revisions is the new assimilation theory presented by Richard Alba and Victor Nee (2003). This concept differs in several important aspects from earlier ones. For example, it does not conceive assimilation as a unidirectional adaption of immigrants, which is apparent in statements such as “the American mainstream has been continually reshaped by the incorporation of new groups” (Alba and Nee 2003, 64). The authors also assert that assimilation is neither an inevitable nor irreversible outcome of immigrant settlement. Moreover, if it occurs, it could do so at different velocities for different groups or individuals and would not necessarily have a strictly linear character.

With the segmented assimilation theory, Portes and Zhou (1993) presented another much- discussed revision of earlier assimilation models. A defining characteristic of their novel approach is a predominant focus on the descendants of migrants, especially their direct offspring, whom they refer to as second-generation.4 Overall, the arguments made by these authors emanate from the observation that the ethnic and cultural composition of the immigrant population in the United States in the early 1990s had changed dramatically compared to the era in which the original assimilation theories were devised. More specifically, the authors state that while only 15 percent of immigrants were of non-European origin in 1940, this trend has reversed after 1960, with only 13 percent coming from Europe. Additionally, the authors estimate that 54 percent of all post-1960 immigrants would be considered non-white according to phenotypical classifications common to the United States (Portes and Zhou 1993, 77-78). Building on this

4 While the terms second-generation immigrant and second-generation migrant are frequently used in public and academic discourse, they are logical contradictions. Since human migration refers to the physical movement and possible settlement of an individual (immigration) it can, by definition, not be a hereditary personal trait. While it cannot be disputed that it is important to study the living conditions, educational trajectories etc. of migrants’ children, it is desirable that this group is not referred to by misleading terms. Such terms might be practical, but they also contribute to the demarcation of in- and out-groups. The present work will, consequently, use these terms only if they were employed by the authors of a discussed text or in an original argument. The decision not to use alternative descriptors in such situations is taken to avoid confusion on the reader’s side.

Theoretical considerations observation, and in line with theoretical considerations by Herbert J. Gans (1992), Portes and Zhou argue that the incorporation of immigrants’ children would not necessarily occur into the socially, politically, and culturally dominant group (i.e., the so-called mainstream) of the CoR. In fact, it would be likely that these individuals would be confronted with high levels of racism and that their social and economic advancement would be hampered by structural disadvantages, resulting in a situation in which „[c]hildren of nonwhite immigrants may not even have the opportunity of gaining access to middle-class white society, no matter how acculturated they become.” (Portes and Zhou 1993, 96). The obstacles observed by the authors coincide with the fact that large parts of the post-1960 immigrant population in the United States settled in inner- city neighbourhoods. These are areas where they, on the one hand, find affordable housing but, on the other hand, live amongst the autochthonous underclass. In this environment, pre-existing ethnically determined subcultures of the disadvantaged youth provide immigrant children with alternative possibilities of identification. According to the authors, these subcultures are constructed in opposition to the perceived mainstream culture which is viewed as discriminatory and, to a large extent, held responsible for the disadvantageous situation in which their members find themselves. Since the devaluation of educational achievements is an important feature of these youth cultures, the incorporation of migrant children into them can lead to their direct assimilation into the aforementioned underclass, leading to a high risk of poverty and unemployment. Besides assimilation in an ethnically determined underclass or the mainstream, Portes and Zhou assume the existence of a third path to incorporation dubbed ‘selective assimilation’. In this setting children of immigrants who are confronted with racism and structural disadvantages would assimilate partially into the CoR (e.g., by respecting legal norms, retaining high educational aspirations, and acquiring the CoR language). At the same time, they would also benefit from co-ethnical networks within the immigrant community established by their parents, to which they retain close cultural and economic bonds (Portes and Zhou 1993, 90). Hence, the important contribution made by the segmented assimilation theory consists of a more thorough conceptualisation of different assimilation paths. However, it has been pointed out that in its application the focus is mostly on the assimilation in disadvantaged minorities, while the other two paths receive very little empirical attention (Vermeulen 2010, 1225, 1227). A controversial area of discussion in migration research is the degree to which this theory could also be applied to other countries, for example in Europe, where ethnically defined and racially discriminated autochthonous minorities do not exist to the same extent as in the United States (for

Theoretical considerations diverging assessments see, e.g., Silberman, Alba, and Fournier 2007, 23; Vermeulen 2010; Schunck 2014, 27).

Despite the strong influence of American migration research on European discourses in this field, the term assimilation is employed much less frequently in the latter. Instead, scholars more often speak of integration in their discussions of immigrant incorporation processes (Lucassen 2006, 17). On the one hand, the European aversion to the term assimilation results, at least partially, from its negative association with historical policies of linguistic and cultural homogenisation. On the other hand, the dislike of the term is connected to the fact that its early versions, as developed by the Chicago School, implied a unidirectional orientation of immigrants towards the dominant cultural norms present in the CoR (Faist and Ulbricht 2014, 14). Nonetheless, differences in the meaning of the terms assimilation and integration in academia are nowadays often perceived as minimal, with some scholars even treating both terms as synonyms (e.g., Brubaker 2001, 540; Vertovec 2009, 78; Stepick and Stepick 2010, 1162; Vermeulen 2010, 1227). However, other authors do maintain that there are differences between them. From a European perspective, Schneider and Crul (2010, 1145) observe that structural aspects are usually taken into greater consideration in analyses which employ concepts using the term integration. While this aspect of their assessment is shared by Alba and Foner (2015b; 2016), who look at different concepts from an American perspective, both groups of authors diverge in their opinion regarding the significance of acculturation in the academic understanding and use of the term integration. Whereas in the opinion of the latter, integration – in contrast to assimilation – “is agnostic about cultural and social change” (Alba and Foner 2015b, 7), the former argue that “with regard to cultural aspects the term integration actually means something pretty similar to ‘assimilation’” (Schneider and Crul 2010, 1145, italics in the original).

An additional observation that can be made is that European scholars usually employ rather comprehensive models to describe the integration of immigrants. A good example of such a holistic approach is a very general definition given by Penninx and Garcés-Mascareñas:

“The term integration refers to the process of settlement, interaction with the host society, and social change that follows immigration. From the moment immigrants arrive in a host society, they must ‘secure a place’ for themselves. […] Migrants must ﬁnd a home, a job and income, schools for their children, and access to health facilities. They must ﬁnd a place in a social and cultural sense as well, as they have to establish cooperation and interaction with other individuals and groups, get to know and use institutions of the host

Theoretical considerations

society, and become recognized and accepted in their cultural speciﬁcity.” (Penninx and Garcés-Mascareñas 2016, 11)

The understanding put forward by these authors points to both the complexity of integration and to the fact that the respective process cannot be equated with the establishment of cultural homogeneity.

To analytically cope with such complexity, integration processes are usually understood and characterised as multidimensional phenomena in current research. That said, there is currently no consensus regarding the number and description of these dimensions. For example, Lacroix, in his comparison of transnational migrants in France and the United Kingdom, differentiates between three dimensions, namely systemic integration, defined as the individuals’ political and economic integration, social integration, and identity integration (Lacroix 2013, 1022f). Other recent studies investigating the relationship of integration and transnationalism differentiate only between two broader dimensions such as structural and sociocultural integration (Snel, Engbersen, and Leerkes 2006, 287; Erdal and Oeppen 2013, 871) or integration in economic or educational terms, and sociocultural integration (Ley 2013, 923). However, these dimensions are often parsed into further subdomains to account for different aspects. Erdal and Oeppen, for instance, name a total of six subdimensions which are based upon the mentioned dichotomy at the higher level (2013, 876). Further theoretical (Esser 2000; Bommes 2012) and empirical (Braun and Müller 2012; Schunck 2014) contributions distinguish between four dimensions of integration.

Based on the described concepts and arguments this work will employ an understanding of integration processes which captures their multidimensional nature. To this end, it will differentiate between the dimensions of cultural, social, structural, and identificational integration. These four domains incorporate the additional aspects employed by some of the aforementioned authors whose concepts include a higher or lower number of dimensions. This approach is also taken to strengthen the comparability of this work with results from earlier studies of immigrants in Europe. An argument against the application of the segmented assimilation theory is the fact that it centres very much on the influence of early live acculturation and socialisation. This, in turn, favours its use in analyses which focus either on migrants who arrived at the CoR at a very young age or on their descendants.

Theoretical considerations

The multidimensional description of integration employed in this analysis will strongly build on the integration model put forward by Hartmut Esser (1980; 2000; 2001) which is highly influential in the German-speaking scientific community. It is important to underline that Esser’s theory is not limited merely to migrant integration but also addresses societal integration and cohesion in general. The integration of migrants and ethnic minorities are indeed special cases which Esser discusses within the general analytical framework he developed (Esser 2001). A key point of Esser’s concept is the distinction between the integration of individuals into an existing social system and the integration (i.e., cohesion) of the respective system itself. To clearly differentiate between these dimensions Esser uses the term ‘social integration’ for the former and ‘system integration’ for the latter. In doing so, he directly builds on David Lockwood (Esser 2001, 3) who posits that “social integration focuses attention upon the orderly or conflictful relationships between the actors […] [while] system integration focuses on the orderly or conflictful relationships between the parts, of a social system.” (Lockwood 1964, 245, italics in the original). Hence, when discussing the relationship between integration and transnationalism, it is social integration at an individual level that becomes the centre of focus, and is usually further divided into several subdimensions. However, before discussing these points in detail, a note regarding the terminology is in order. As most of migration research concentrates on integration efforts and effects on an individual level, system integration is seldom mentioned explicitly. Consequently, the need to differentiate it from social integration in Lockwood’s and Esser’s sense does not present itself. This leads to a situation in which the term social integration is commonly used not to denote one of two general types of integration, but as a label of just one particular dimension of integration on an individual level (e.g., Snel, Engbersen, and Leerkes 2006; Braun and Müller 2012; Erdal and Oeppen 2013; Lacroix 2013). This specific subdimension is called ‘interaction’ in Esser’s model. Nevertheless, the present work will follow the trend in international research to employ the term social integration as a descriptor of this specific aspect of incorporation at an individual level. This decision is moreover taken to facilitate the comparison of this analysis with other empirical research on the relation between transnationalism and integration.

Cultural integration refers to what Esser describes as ‘culturation’ (Kulturation, Esser 2000, 272) as it concerns the knowledge and competencies required by individuals in order to act successfully in varying social contexts of the CoR. A relevant distinction with regard to

Theoretical considerations culturation refers to the time and circumstances in which it takes place. As Esser explains, ‘enculturation’ is the first of two processes which can be distinguished (Esser 2001, 9). It describes the acquisition of cultural knowledge as part of an individual’s early socialisation. On the other hand, ‘acculturation’ describes culturation processes in a later stage of an individual’s life in cultural surroundings in which they did not grow up. The author underlines, that, especially regarding the acquisition of languages, this process is the more successful, the earlier in life it starts. Hence, the successful acquisition of a language relies on early exposure, which, in turn, points to the significance of the age at which individuals migrate to the CoR. This assessment is shared by John Berry (1997) who looks at the adaption of migrants to new cultural and social settings from a psychological point of view. Referring to individuals who migrate at a later point in their life and, therefore have been acculturated in another cultural context, he notes that “a whole life in one cultural setting cannot easily be ignored when one is attempting to live in a new setting” (Berry 1997, 22).

From a general perspective, Esser argues that the knowledge and capacities acquired during acculturation could and should be seen as a form of human capital (Esser 2001, 8). Such capital can be used by the specific actor to reach a certain objective in an interaction with other individuals within the same system. Similar to Esser, other authors hint to the importance of the acquisition of the CoR language as a major part of cultural integration (Braun and Müller 2012, 265; Schunck 2014, 258). This is quite logical since successful interactions with members of a social group are far more unlikely if there is no common way of communication. However, the success of communication is not necessarily and not only determined by the command of vocabulary and grammatical structures. It is in fact also important to know the underlying cultural reality of a language to enhance the probability that the recipient of a message will be able to decipher it correctly and interpret it in the way intended by the speaker. This point is eloquently explained by Alfred Schütz in a short summary of the relevant elements of face-to-face communication:

"There are first the words uttered in the meaning they have according to dictionary and grammar in the language used plus the additional fringes they receive from the context of the speech and the supervening connotations originating in the particular circumstances of the speaker. There is, furthermore, the inflection of the speaker’s voice, his facial expression, the gestures which accompany his talking." (Schütz 1945, 543-44)

Theoretical considerations

Michael Bommes calls this dimension cognitive integration and points out that it also comprises the individual’s familiarisation with behavioural and situational patterns (Bommes 2012, 114). The observations of Schütz and Bommes are important given that a lack of knowledge in this area often causes what is commonly known as cultural misunderstandings. This refers, for example, to gestures which accompany verbal communication but may have starkly different or even opposite meanings across cultures. The same holds true for typical behaviour in certain interpersonal situations. What might be regarded as basic courtesy in one culture could constitute an insult in another. This directly points to the acquisition of normative knowledge which is another element of this dimension of integration (Esser 2000, 272; Bommes 2012, 114).

Considering the fact that the nuances of a culture are most easily internalised through prolonged direct contact, it can be assumed that not only the age at migration but also the time spent in a particular CoR is important regarding cultural integration. This assumption is in line with Esser’s (2001) assessment that successful acculturation is highly dependent on opportunities and regular exposure to the target culture. In the literature such exposure is mainly discussed with regard to the physical and social environment migrants live in. Nevertheless, it could be assumed that general knowledge of other parts of the CoR (i.e., regions in which the individual is not currently residing) would bring migrants in contact with different aspects of its culture and, therefore, also enhance their integration in the cultural sphere.

Finally, it has been pointed out by Esser (2001, 25) and Berry (1997, 23) that the course of the acculturation processes also depends on the cultural distance between the countries of origin and of residence. Berry gives differences in language and religion as examples, and concludes that adapting to new social and cultural settings is the more difficult for migrants, the greater the dissimilarities between their contexts of origin and residence.

Alaminos and Santacreu summarise social integration as the dimension which “deals with participation, through meaningful social relations” (2009, 109). This means it builds on the interactions of migrants with natives of the CoR and, consequently, is also related to cultural integration. The array of those social interactions is not limited to certain social fields or circumstances. In this sense, the term corresponds to the individual’s integration into groups which can be situated in a private or a professional context. However, in the latter, it is important to stress the social aspect as opposed to the structural. Hence, this dimension does not refer to the

Theoretical considerations question of whether or not the individual has acquired a specific position within an enterprise or the labour market in general. While not limited to this aspect, it does, however, refer to the degree in which individuals are connected and acquainted with other persons in the workplace and whether they are part of informal circles and activities which might spill over into their private lives.

Regarding the private sphere, aspects investigated within this dimension of integration include friendships, membership in organisations and the general (ethnic) composition of social networks (Bommes 2012; Erdal and Oeppen 2013). It is in this context that Esser asserts a small difference in the character of cultural and social integration. While the former – according to him – has a somewhat ‘technical character’ the latter involves emotions to a greater extent (Esser 2001, 11). Esser also emphasises that social interaction depends on opportunities much in the same way as acculturation does. For example, if the housing market in a city is strongly segregated and migrants regularly live in districts separated from the native population, social interaction requires a high motivation on the side of the migrants or the CoR natives. On the contrary, in a comparatively heterogeneous neighbourhood, the likeliness of intercultural contacts is much higher, requiring far less individual motivation.

Finally, the bidirectional interrelation of social and cultural integration should be mentioned (Esser 2001; Bommes 2012, 114-15). On the one hand, social integration thrives on the practical use of knowledge acquired from acculturation processes. On the other hand, migrants with more frequent social interactions are more likely to encounter opportunities to deepen their understanding of relevant social mechanisms and cultural aspects in their new country. Thereby, social integration also presents the individuals with opportunities to acquire additional cultural capital.

The dimension of structural integration describes the positioning of an individual within social systems and organisations as well as its access to and acquisition of resources such as income, education, and reputation (Bommes 2012, 114; Erdal and Oeppen 2013, 876). Hence, structural integration takes place when an individual takes a pre-existing position within a system and thereby becomes part of it. A notable example would be an individuals’ participation in the labour market of the CoR which, at the same time, indicates its integration in this system. Both, reputation and available income, to take the examples of social and financial resources, depend

Theoretical considerations highly on the degree of integration achieved by the migrant. Unsurprisingly, this type of integration possesses certain preconditions. This means that usually not every individual can take any position within a system just because they desire to do so. On the part of the individual, it often requires certain competencies. The specific requirements in formal education for a particular professional position are certainly the most common example. However, as Esser stresses, the absence of discrimination – i.e., a certain openness of a system – is another important precondition of successful structural integration (Esser 2000, 272).

There is an agreement in the literature that the bestowal of rights constitutes another key aspect of structural integration. A frequently given example is the naturalisation of immigrants (e.g., Bommes 2012, 114; Braun and Müller 2012, 265; Erdal and Oeppen 2013, 876). Naturalisation permits individuals to participate politically in the CoR and can also have very practical consequences on the international mobility options resulting from visa requirements for holders of different passports.

Several authors (Esser 2001, 9; Snel, Engbersen, and Leerkes 2006, 287; Bommes 2012, 114) highlight the fact that structural integration is strongly connected to the previously discussed dimensions. Individuals are able to take certain positions because of their successful culturation, e.g., in terms of language knowledge. Furthermore, through placement on a certain position individuals might come into contact with more people, hence intensifying their level of social interactions. This might, in turn, favour further culturation and strengthen their identification with different social circles in the CoR and its culture.

This directly points to the remaining dimension of identificational integration which concerns “the claims of belonging and identity made by migrants themselves” (Bommes 2012, 114). Identification can be understood as a process in which the individual develops an emotional relationship with a system. This, in turn, leads to a situation in which the individual actors see themselves as part of a collective that shares some kind of group identity (Esser 2000, 275). In the context of international migration, identification with the country of immigration would be an obvious example. In this case, migrants could start to perceive themselves as part of what

Theoretical considerations

Benedict Anderson described eloquently as the ‘imagined community’ (Anderson 2006) of the CoR.5

According to Esser, different forms of identificational integration are possible. These range from a value based identification, associated with a high intra-group solidarity, to the integration through nothing more than compliance with rules in cases where individuals feel that any attempt to change the system would be futile and therefore accept it as it is (Esser 2001, 12-15). Braun and Müller also hint to the fact that there are different levels of identificational integration by stating that it “consists in a strong feeling of belongingness or at least the acceptance of the values of a social system.” (2012, 265).

Besides the differentiation between the above-described dimensions of integration, an important feature of Esser’s concept is the proposal of four ideal types of migrant integration which are possible in each of them. These are: multiple integration, assimilation, segmentation and marginalisation (2001, 19-20). In this categorisation, multiple integration describes a situation in which a migrant is integrated in more than one context. In his original argument, Esser refers to the simultaneous integration of immigrants in a co-ethnic milieu in the CoR and in the host society. In contrast, assimilation describes a situation in which migrants are only integrated in the host society while all significant ties to the co-ethnic milieu are terminated (or not established in the first place). The opposite constellation is segmentation in which the individual is only integrated in the co-ethnic milieu without any meaningful participation in the mainstream society of the CoR. The fourth ideal type, marginalisation, indicates the absence of integration in either context. These four types of integration are sometimes also described as outcomes of integration processes (e.g., Schunck 2014, 33). While this is not an unreasonable interpretation, it should be kept in mind that the terms, in this case, are used to describe a certain status at a specific point in time. Hence, it would be more precise, albeit cumbersome, to speak of outcomes at the time of observation. What is meant by this is that these outcomes are not static endpoints of a process. This is due to the fact that social processes and cultures alike are in permanent flux. Consequently, an individual which is assimilated in a specific dimension today might be marginalised in the same dimension ten years from now.

5 Anderson coined the term not with respect to migrants but as part of his general analysis of the emergence of nationalism.

Theoretical considerations

Another important consideration concerns the use of the term assimilation. In the assimilation models mentioned at the beginning of this chapter, the term is used in a global sense to describe a process and its possible outcome. Hence, the four ideal types could also be understood as different forms of assimilation. Multiple integration would then correspond to assimilation in various contexts, while assimilation in Esser’s sense could be seen as full-assimilation in the CoR context. In contrast, in Esser’s model the process of becoming part of the CoR and participating in it – at a general level – is described as integration, with assimilation being one of four ideal- typical forms in which it can occur or how its outcome at a specific point in time can be described. In analytical parlance assimilation is understood in the present work not as the variable of interest itself, but rather a value it can take. Furthermore, it is important to note that the mentioned ideal types in principle apply to all four dimensions of integration individually. This means a migrant could, for example, show multiple social integration while being structurally assimilated in the CoR.

However, while this classification offers two modes of integration which would indicate successful incorporation in the CoR, Esser argues that especially in the social realm, multiple integration, though theoretically possible, would be empirically infrequent and, at best, limited to children of diplomats, academics, and artists (Esser 2001, 21; Esser 2003, 8-9). This leads him to conclude that, besides such rare exceptions, successful social integration could only be achieved in the form of assimilation. Consequently, assimilation occupies a very prominent position in most of Esser’s theoretical and empirical work. In this context it should not go unmentioned that Esser assumed in his earlier works that assimilation in the four dimensions would occur in a specific order (Esser 1980, 231), much in the same sense assimilation theorists of the Chicago school did. Specifically, he assumed that cultural assimilation would be followed by structural, social, and identificational assimilation. However, this assumption is not present in Esser’s more recent texts. This important modification, or omission, is also mentioned by Schunck (2014, 32).6

6 In general, the idea of a consecutive order is not very convincing. First, because it presupposes a linear process of integration. Contrary to that it has already been mentioned that the respective processes should instead be conceived as dynamic since a specific situation (e.g., assimilation in a dimension) can not only increase but also decrease over time. Second, it is not apparent why, for example, identificational assimilation should only occur once individuals are assimilated in all other dimensions. On the contrary, it is, e.g., not hard to imagine that refugees could identify very strongly with their CoR after only a short period even though they might have little social contact to the country’s population and might not speak its language.

Theoretical considerations

As discussed, Esser’s original argument regarding social integration builds on the dichotomy of the mainstream society in a CoR and an ethnic community in the same country. However, as will be shown below, empirical data indicate that a relevant share of immigrants retains social and material cross-border connections to their CoO after settlement in the CoR. Furthermore, research findings show that frequent cross-border practices are not limited to migrants but that individuals without personal migration background also engage in such activities and pertain to border- crossing social networks (Mau 2007). Therefore, instead of concentrating on the settlement context, this study will define cross-border relationships and the transnational space constituted by them as the second social system in which migrants can integrate. Indeed, Esser indicates in a marginal note in one of his papers (2003, 7) that this would constitute a feasible adaption of his theory. In this framework, aspects of transnationalism could be equated with integration in a border-crossing social system. The ideal types could then be used to assess the relation between transnationalism and integration in the CoR. Following on from Esser’s argument that multiple integration is not a viable option for most individuals, a clear negative relationship should exist between indicators of transnationalism and indicators of integration. Such an effect would mean that transnationalism, in fact, fosters segregation, which would leave assimilation as the only successful path to integration. However, if no relations or positive associations between transnationalism and integration are visible, it could be concluded that both phenomena do not impede each other but might even be positively related. Hence, the absence of negative links would indicate the possibility of multiple integration.

2.2.3 The relevance of context factors for the integration process Existing research leaves no doubt that the context in which migrants live, work and interact with others is highly relevant to their integration process (e.g., Portes and Zhou 1993; Esser 2001). Important elements to consider are the institutional framework of the labour market and the educational system, the housing situation, and the CoR’s legal system, which might ease or complicate migrants’ participation in relevant social systems. Crul and Schneider (2010) have pointed out that many of these aspects and their effects differ not only between countries but also between cities within the boundaries of nation states. In their study, they compare individuals who live and were born in one of eight European Union member-states and have at least one parent from Morocco, Turkey or former Yugoslavia.7 They find that even children of migrants

7 Details regarding the sample composition and data are given in Groenewold and Lessard-Phillips (2012).

Theoretical considerations who share the same ethnic background differ greatly between these countries. Differences are shown, for example, in the degree to which they are involved in their neighbourhoods (e.g., feeling responsible for and comfortable in it), the ethnical composition of their social networks, and their national identification. Crul and Schneider argue that these differences are strongly influenced by differing integration contexts and identify a number of relevant factors such as educational systems, labour market regulations, and societal diversity (Crul and Schneider 2010, 1257). The authors further find that context factors differ not only between but also within CoRs.

While factors such as discrimination might often influence the life circumstances and chances of migrants, it should be stressed that context factors can also be of a positive nature or at least have their origin in positive – albeit not necessarily altruistic – intentions. Social programs offered by governmental and non-governmental agencies directed at migrants or the general population would be a case in point. Given these considerations, this work not only compares migrants in different CoRs but also from two distinct CoOs. As only one of these CoOs is within the European Union (which all the CoRs are), the individuals in both origin groups differ with respect to the rights assigned to them due to their nationality and irrespective of their individual titles of residence.

2.3 Transnationalism in migration research

2.3.1 Defining the phenomenon The concept of transnationalism rose to prominence in migration research during the last decade of the 20th century (Portes, Guarnizo, and Haller 2002, 279; Morawska 2003, 611; Levitt and Glick Schiller 2004,1004). It owes its success to the empirical observation that a relevant share of migrants tends to stay connected to their CoO after settling abroad. Furthermore, migrants might not just sustain existing cross-border relations but establish new ones after leaving (Faist, Fauser, and Reisenauer 2013, 6). Hence, the term transnationalism in this context refers to activities and networks which transcend international borders and thereby link individuals and the organisations they may form in various states. Such links include social relationships, economic undertakings, political activism and participation, as well as individual mobility and cultural activities (Vertovec 2009, 9; Kuhn 2015, 31; Pries 2008, 44). These practices and connections are, furthermore, characterised by a certain degree of regularity and consistency over time (Portes 1999, 464; Portes, Guarnizo, and Landolt 1999, 219; Morawska 2002, 134; Pries 2008, 45).

Theoretical considerations

Additionally, as will be argued below, it is beneficial to consider the possession of knowledge and skills which enhance the probability that individuals will engage in transnational activities or be part of transnational networks, in the basic understanding of transnationalism (see Kuhn 2011, 814).

While the concept of transnationalism is also used in other scientific fields (e.g., in economics, see Pries 2008, 182-88), it became particularly popular in migration research. The respective studies investigate different types of transnational activities and aspects of cross-border life involving migrants (or their offspring) and their CoO. They focus, for example, on political transnationalism and dual citizenship (e.g., Bauböck 2003; Østergaard-Nielsen 2003; Itzigsohn and Villacrés 2008), migrants’ economic activities, or on migrants and migrant organisations as actors in the development of the CoO (e.g., Portes 2009; Escobar 2010). The understanding of transnationalism and transnational interactions in this strand of research consequently does not only relate to migrants themselves but also to their friends and family in the CoO and CoR, and other sedentary individuals who might be (indirectly) affected by the transnational activities that migrants engage in (Levitt 2001, 198; Glick Schiller 2005, 28; Bauböck 2006, 28).

If cross-border contacts are sustained over time and result in repeated transactions between the involved individuals, they can lead to the establishment of social structures which have, for example, been described as transnational social fields (Glick Schiller, Basch, and Blanc-Szaton 1992a; Portes, Guarnizo, and Landolt 1999; Levitt 2002) or transnational social spaces (Faist 2000; Pries 2008). Faist emphasises that these spaces are not to be equated with communities or groups. Instead, they should be understood as “relatively stable, lasting and dense sets of ties reaching beyond and across the borders of sovereign states.“ (Faist 2010, 13). He goes on to specify that these ties reflect the repeated interactions of individuals in two or more geographically distinct places. Consequently, such spaces provide individuals with opportunity structures that do not only correspond to those inherent to the distinct geographic locations but also to additional opportunities that arise from their connection. However, the given definition also infers that it is not inevitable that transnational activities would lead to the formation of transnational social spaces, as these formations are characterised by a certain degree of stability. In general, the literature asserts that transnationalism has a process-like character (Glick Schiller, Basch, and Szanton Blanc 1999; Portes 2007, 78; Kuhn 2015, 24). This idea is emphasised by Faist et al. who argue that “transnational social spaces are not static but […] dynamic social

Theoretical considerations processes” (Faist, Fauser, and Reisenauer 2013, 55; see also Faist 2010, 13). Consequently, the aforementioned stability mostly refers to the interpersonal links between actors, while the nature of activities performed within this space (e.g., communication, visits, exchange of material goods) might vary over time. Furthermore, it should be noted that its character as a process does not mean that transnationalism should be understood as a target oriented event. In fact, the term highlights the fluid non-static nature of this phenomenon.

While the aforementioned research pertains to transnational activities and connections of migrants, researchers investigating transnationalism on an individual level have only recently gone beyond the realm of migration studies to focus on the cross-border activities and networks of non-migrant populations (e.g., Mau 2007; Dahinden 2009; Kuhn 2015). The results of a survey on the transnationalisation of social relations amongst the general German population conducted by Steffen Mau and colleagues in 2006 are most informative in this regard (Mau 2007; Mau and Mewes 2007; Mau, Mewes, and Zimmermann 2008). The assumption underlying this research was that a fundamental change in the structure of interactions and networks of the German public occurred during the last decades of the 20th century. According to the findings of this project, this change led to a situation in which social relations and daily life of the German population increasingly transcended international borders (Mau 2007, 9). Mau characterises this as a surge of de-nationalisation in social and daily life that manifests itself, for example, in the rise of transnational marriages, the extension of spaces of interaction across borders, and the decreasing importance of the nation state as primary unit of identification (Mau 2007, 10-11). In their analyses, Mau and collaborators show that indeed a substantive part of the German population engages in transnational activities. An essential contribution of these authors is their mapping of the relevant transnational spaces (see Mau (2007), chapter 11-15 or Mau and Mewes (2007) for a less extensive discussion). Since their focus is not on immigrants the question of the geographic scope of these transnational spaces is by default an open one which distinguishes their research from the vast majority of migration studies. While various authors mention in passing that transnational activities and networks of migrants can go beyond the dichotomy of CoR and CoO (e.g., Guarnizo and Smith 1998, 13; Vertovec 2009, 13; Faist, Fauser, and Reisenauer 2013, 54), most empirical studies do not investigate the degree to which places and individuals in third countries are part of migrants’ transnational experiences.

Theoretical considerations

This focus on transnational activities and networks between the CoO and CoR might not only lead to a situation in which the degree of migrant transnationalism is underestimated but it also renders some of the conclusions which are drawn from such data questionable. The interesting and informative study of migrants from Afghanistan, Burundi, Ethiopia and Morocco living in the Netherlands by Özge Bilgili (2014a) illustrates this point. In this analysis, Bilgili focusses on the relation between sociocultural integration and different sociocultural transnational activities. Regarding the latter, she limits her analysis to practices which are directed at the respondents’ CoO. Her analysis reveals that Burundians and Afghans have significantly less contact with family and friends in their CoO than individuals from the other two countries. Bilgili argues that these differences might be partly due to the political and social situations in Afghanistan and Burundi which might have caused not only the participants of the study but also parts of their families to emigrate. Consequently, those respondents may have fewer remaining contacts left in their CoOs, compared to migrants from Morocco or Ethiopia (Bilgili 2014a, 297). The author furthermore reports that Afghans and Burundians are also much more likely to consume CoO related media and art. This leads her to the conclusion that this media consumption may serve as “a substitute for social contacts with family and friends.” (Bilgili 2014a, 299). It is not the intention to question this thesis altogether, but an obvious weakness arises from Bilgili’s own argument that a high share of family members and friends from these countries might have migrated themselves. This could explain the lower level of transnational communication with the CoO. However, this does not automatically indicate fewer transnational social contacts in general. On the contrary, it is possible that the transnational networks of Burundian and Afghan migrants are simply more complex regarding their geographic structure than those of Moroccans and Ethiopians. Hence, Bilgili’s argument could have benefited greatly from a geographically more diverse scope which included transnational activities and ties directed towards third countries.

As indicated at the beginning of this chapter, the understanding of transnationalism applied in this work does not intend to imply that nation states are unimportant in the context of modern-day migratory phenomena. The definition of transnational activities as practices which transcend borders and connect individuals in different states already indicates that modern states are not rendered obsolete (Guarnizo and Smith 1998, 7; Dahinden 2009, 1383; Pries 2010, 15-16). In fact, transnational practices are still “anchored” (Vertovec 2009, 12) in places and thereby

Theoretical considerations influenced by the legal and cultural frameworks which govern them. Authors such as Guarnizo and Smith (1998, 13), and Faist (2000, 200) highlight how physical locations that are connected by transnational networks and activities exert an important influence on these practices. Relevant factors are, for example, the legal framework of the respective countries, the existence of differently sized groups of co-nationals, migrant organisations and the public opinion towards immigrants. The “structuring capacity of the nation state” (Kuhn 2015, 24) is particularly visible where migrants demand membership rights in more than one political entity. In this sense, Bauböck argues that „[t]ransnational citizenship is […] still about migrants’ affiliations with distinct and clearly bounded political communities.“ (2006, 29). Such affiliations might blur the boundaries of political entities, but they do not dissolve them. The same holds true for activities on a number of different levels. If for example, money earned in a migrant’s CoR is invested in his or her CoO, the ability to do so on an ongoing basis is likely influenced by the development of income levels in the former. Hence, changes in taxation and social benefit deductions in one state will directly influence economic activities in the other. In the same sense income earned through investments in the CoO are – with regard to tax exemption limits – usually of interest to tax authorities in the CoR.

2.3.2 Migrant transnationalism in historical perspective The question of whether transnationalism is indeed a new phenomenon is an important one within migration research. To shed light on this point, it is helpful to reflect on the reasons why transnationalism occurs in the first place. A basic assumption which can be made in this regard is that not all people who decide to leave their home community to live in a foreign country will want to cut all ties with the former. As underlined by Escobar (2010, 106) contemporary migration is to a large extent the consequence of economic differences and imbalances between sending and receiving regions in the context of globalisation. Consequently, it can be assumed that migration is more likely to be driven by rational considerations than by an emotional distance to or rejection of the CoO, its population or culture. However, this argument refers only to migration that occurs voluntarily. There are, unfortunately, a number of reasons, such as political persecution on an individual or group level, discrimination, natural disasters, and war, which might force people to migrate and seek refuge elsewhere. Consequently, it is important to differentiate between forced displacement and migration which occurs not because of an immediate threat to an individuals’ life or wellbeing. According to the International Organization

Theoretical considerations for Migration (2016, 5, 8), the number of international refugees and asylum seekers reached a distressing level of 24.5 million individuals in 2015. However, they are still a clear minority amongst the world’s 244 million international migrants. For those who do not migrate from regions where large-scale displacement is occurring it can be assumed, that a majority leave behind at least some individuals who are important to them and in whose fate they are invested. This leads to a situation in which migrants settle in another country while remaining connected by social relations and various activities to their CoO. However, such ongoing connections are not entirely new. Historical connections kept by migrants with people in their regions of origin have long been a subject of systematic research (for overviews see Morawska 2003; Foner 2007). That said, historical evidence of enduring connections between a person’s region of origin and settlement in the context of migration is not a convincing argument against the concept of transnationalism as a whole. From a methodological point of view, it has been argued that research which now identifies patterns of transnationalism in historical migration movements actually supports the value of the approach. In such cases, this new ‘analytical lens’ would allow scholars to identify and describe phenomena which had eluded them before or which were disregarded (Portes 2003, 874-75; R. C. Smith 2003, 725). This argument underlines an important point: the term transnationalism can be simultaneously used to describe a set of activities and related phenomena (such as networks), and to depict a specific analytical focus. Boccagni (2012, 119) points to this duality as a potential source of confusion. Hence, the transnational research perspective in migration studies, which the notion of a new analytical lens is referring to, is characterised by increased attention to cross-border ties and activities of individuals, leaving behind the sole focus on either CoR or CoO. This conscious extension of the research perspective may very well have led to the aforementioned re-evaluation of historical phenomena. Conversely, individual transnationalism refers to ties and social practices through which people interact across international borders with each other, with organisations, and with institutions. It is this engagement for which it can indeed be argued from an empirical point of view, that it shows a number of characteristics which distinguish it from historical patterns (Portes 2003; Foner 2007).

Most prominent among these characteristics is the use of new, fast and cheap means of communication. These are, first, internet-based technologies, such as email, Voice over Internet Protocol (VoIP) communication, video chat, and social networking sites (e.g., Facebook, Hi5,

Theoretical considerations

Google+). Second, the spread of mobile phones and the increasing use of instant communication applications have to be mentioned. The latter is a more recent occurrence and refers to services which allow a two-way communication (chat) such as WhatsApp, and to services which enable the sender to communicate news and information without necessarily engaging in a dialogue with message recipients (e.g., Twitter). Finally, Vertovec (2009, 57) underlines the importance of the spread of calling cards for international telephone calls and the drop in the respective costs. While long-distance communication was possible before, it has reached a new level of immediateness. As Faist and co-authors put it: “In contrast to the past, contemporary migrants do not have to wait two weeks or even longer for a reply to a letter but can communicate immediately with their relatives and friends abroad” (Faist, Fauser, and Reisenauer 2013, 51)

Another aspect which characterises modern-day transnationalism is the availability of long- distance transportation at comparatively low prices (Pries 1996, 467; Escobar 2010, 106). These possibilities are not limited to air travel. Within the same continent, and in the case of adjoining continents, long-distance coaches, trains, and individual mobility by car are important, too. This does not mean that the majority of migrants could economically afford to go back and forth between CoO and CoR at any time. Nevertheless, it does mean that in most cases even cross- Atlantic or cross-Pacific migration does not have the character of a one-way trip anymore. Again, the argument here is not that circular movement is something entirely new,8 but that, due to its affordability and the considerable cut in travel time, it has become a realistic option for a much larger share of migrants.

Furthermore, modern transnationalism is also characterised by unprecedented economic connections between migrants and their CoOs. Migrants send money, invest in companies – or even set them up – and purchase property back in their CoO (De Haas 2005). The development of worldwide private money transfers to other countries highlights this phenomenon. According to the World Bank (2015a), the total value of international personal transfers and cross-national compensation of employees rose from $1.4 billion in 1970 to $30.2 billion by 1980. This value has roughly doubled within each of the following two decades, reaching $118.2 billion in 2000. However, the increase in the years after was even more extreme. Until the outset of the economic crisis in 2008, the value grew to $311.2 billion and only decreased slightly to $293 billion in

8 Foner (2007, 2484) mentions, for example, that already in the early 20th century some Italians regularly travelled between the United States and Italy for work.

Theoretical considerations

2010. An all-time high was reached in 2013 with $365.2 billion. While a considerable drop could be observed by 2014 ($310.5 billion), this value still meant that money transfers had grown approximately by a factor of 221 since 1970 and by more than 162 percent since the millennium alone. It is necessary to underline that these impressive figures do, however, also include money transfers of sedentary individuals and salaries paid in the context of trans-border working arrangements.9 Nonetheless, these figures highlight the aforementioned trend as it can be assumed that the main share of this growth in private international payments can be attributed to migrant remittances. Furthermore, it has repeatedly been pointed out that monetary remittances made by emigrants have become a primary source of external revenues in a number of countries (e.g., Itzigsohn 1995; Gosh 2006, 55; Escobar 2007, 49). Recent World Bank data show the global character of this phenomenon. For example, in 2013 remittances corresponded to 21 percent of the GDP of Armenia, 18 percent in Lebanon, and 10 percent in Guatemala. However, the importance of such transfers is not limited to smaller countries and economies: remittances also equalled a considerable share of the GDP in some bigger countries, such as Hungary (three percent), Romania (two percent) and Mexico (two percent) (World Bank 2015b).

Another sphere in which a previously unknown level of interconnectedness between migrants and their CoOs can be observed is the political arena. In recent decades emigrants all over the world have engaged in political matters in their CoOs, while simultaneously demanding the extension of political rights to non-resident citizens. In light of the growing economic importance of migrant remittances, a rising number of CoO governments are willing to meet some of these demands, at least at a symbolic level. Most prominent amongst such measures is the constitutional implementation of the right to hold more than one nationality (for examples see: Mazzolari 2005, 6; Pötzschke 2016, 163). Some countries also introduced measures which restored voting rights in national elections to non-resident citizens and allowed them to cast their votes from abroad. One of the more recent examples is Turkey, whose emigrants were able to vote from abroad for the president10 and parliament for the first time in 2014 and 2015, respectively (Okyay 2014; Elger and Gezer 2014; Migazin 2015). Before this, Turkish citizens

9 These figures correspond to what the World Bank defines as ‘personal remittances’. These are “personal transfers and compensation of employees. Personal transfers consist of all current transfers in cash or in kind made or received by resident households to or from nonresident households. […] Compensation of employees refers to the income of […] workers who are employed in an economy where they are not resident and of residents employed by nonresident entities.” (World Bank 2015a). 10 It should be noted that the Turkish president was elected in 2014 for the first time by popular vote.

Theoretical considerations who resided abroad had to return to their home region or, alternatively, cast their votes at polling stations on the Turkish border (Okyay 2015, 119, 143). Some countries, such as Italy and Colombia, have even gone a step further and introduced direct parliamentary representation of emigrants through the establishment of additional designated mandates in their national parliaments (Tager 2006, 38; Escobar 2007; Pötzschke 2016).

2.3.3 Systematising transnationalism Numerous scholars have contributed to efforts to deepen our understanding of transnationalism by systemising it. These efforts can be divided into two main approaches which are the classification of transnationalism’s general nature, and the description of transnationalism as a multidimensional phenomenon.

Properties repeatedly used to describe the nature of transnationalism are the frequency in which transnational activities are performed and their degree of institutionalisation. Itzigsohn et al. (1999) distinguish in this sense between ‘narrow’ and ‘broad transnationalism’. The former is used to describe situations in which people engage in transnational activities on a regular basis and within a rather formalised setting. In contrast, the second category characterises practices that are infrequently performed and lack a high level of institutionalisation. Levitt’s concept, which distinguishes ‘core’ and ‘expanded transnationalism’ (Levitt 2001), is another example using the frequency of activities as a main distinguishing feature, building – among others – on the observations made by Itzigsohn and his co-authors.

In their analysis of transnationalism’s nature, Guarnizo and Smith (1998, 13) see power relations in the transnational social field as a defining characteristic that differentiates between ‘transnationalism from above’ and ‘transnationalism from below’. The efforts of CoOs to extend their influence on emigrants through their (symbolic) inclusion in the nation by granting extraterritorial civic rights, accepting dual citizenship, or establishing transnational institutions, all constitute examples of transnationalism from above. As the authors stress, it is, nonetheless, not only state actors or global acting entities which can be the driving force of transnationalism from above. Emigrants could also be perceived by the population of the city or region of origin as actors of transnationalism from above when investing financially in their CoO.

Another influential classification of transnationalism has been proposed by Itzigsohn and Giorguli Saucedo (2002), who focus in this case on specific determinants of cross-border

Theoretical considerations engagement. The authors developed this approach while examining the relation between transnationalism and integration; consequently, this systematisation is situated at the intersection of the two concepts. In brief, they propose three ideal types to characterise transnationalism. These are: ‘linear transnationalism’, ‘resource dependent transnationalism’ and ‘reactive transnationalism’ (Itzigsohn and Giorguli Saucedo 2002, 771-72, 781). In the first case the authors argue that a continuously strong identification with the CoO and, more importantly, the preservation of strong social links to this country can lead to the engagement in transnational activities which basically serves to sustain pre-existing relations and to fulfil lasting obligations (e.g., sending money to family members who care for the migrants’ children). As the name suggests, the underlying hypothesis of resource dependent transnationalism is that migrants are particularly likely to engage in transnational activities if they are in an economically stable situation. This characterisation rests on the assumption that the more comfortable an individual’s financial situation is, e.g., due to a higher degree of labour market integration, the more likely they are to be able to spare resources which can be invested in transnational activities such as visits to or business investments in the CoO. Lastly, reactive transnationalism takes into account the influence that particular contextual factors in the CoR exert on transnational engagement. More specifically, the core assumption is that individuals might turn to transnational networks and activities in the face of discrimination and social marginalisation experienced in the CoR. In line with this argument, Faist (2000, 198) lists discrimination as one of several aspects which can contribute to the stabilisation of transnational spaces.11

Generally speaking, the examination of factors which determine transnational activities is extremely important in understanding the possible relation between transnationalism and integration, since the same aspects might have different effects on both phenomena. Amongst the most often studied influencing factors are the time migrants have already spent in the target country (see for example Portes, Guarnizo, and Haller 2002; Snel, Engbersen, and Leerkes 2006; Ley 2013; Schunck 2014), gender (e.g., Guarnizo, Portes, and Haller 2003; Itzigsohn and Giorguli Saucedo 2002; Dekker and Siegel 2013; Carling and Hoelscher 2013), and education (e.g., Guarnizo, Portes, and Haller 2003; O'Flaherty, Skrbis, and Tranter 2007; Schans 2009).

However, while there is a consensus in large parts of migration research that transnationalism is an important feature of current migratory movements, this should not be mistaken as a claim that

11 Further typologies have been proposed by different authors. An overview is provided by Vertovec (2009, 18).

Theoretical considerations all migrants are involved in transnational activities. On the contrary, it is likely that the degree to which individuals participate in transnational spaces and develop or perform transnational practices varies strongly (Guarnizo, Portes, and Haller 2003, 1238; Vertovec 2007, 1043; Dahinden 2009, 1366). The degree of transnational engagement is also influenced by a large number of factors which differ in the case of every single migrant. Some of these are the individual’s migration biography, his or her legal status, the locality of settlement in the CoR, the political situation in the CoO, and the economic and cultural capital available to the migrant (Vertovec 2007, 1043). Morawska’s (2002) study is enlightening in this regard, as the author details differences in transnational engagement of various groups in the United States of America. She shows that Chinese migrants in New York City are much less transnationally active than immigrants from Jamaica and the Dominican Republic who settle in the metropolis. According to the author, this is largely due to the dominant migration pattern amongst the Chinese which is characterised by irregular entry into the country and generally associated with high costs. Consequently, migrants in this group often have an uncertain legal status and find themselves in (at least temporary) economic dependency on money lenders. In this context possibilities for sending money back to the CoO or to travel there in person are limited. Of course, such differences can not only exist between migrants from different countries but also within a national group.

As will be shown below, the empirical assessment of determinants of transnationalism varies considerably between studies. However, with regard to the partially different conclusions drawn in the field, it has to be taken into account that even scholars who nominally investigate the same facets of transnationalism often differ in their operationalisation of specific phenomena. Differences in operationalisation are also relevant in regard to control variables included by the authors. However, the literature is not only diverse with respect to operationalisations but also regarding the aspects of transnationalism that are taken into consideration (e.g., transnational mobility, transnational communication behaviour, and remittances). This is relevant in so far as previous research has shown that various aspects of transnationalism are influenced in different ways by the same factors (see, e.g., Schans 2009, 1179; Schunck 2014, 251). Hence, disparities are not only unsurprising but should be expected. This aspect leads to the above-mentioned second strand of arguments, which proposes systematising transnationalism by differentiating between several dimensions of the phenomenon.

Theoretical considerations

Vertovec intends, for example, to highlight “the most signiﬁcant institutions and aspects of social life affected by transnational ties” (Vertovec 2009, 61). To this end, he differentiates between aspects of transnationalism concerning families, gender, habitus, cultural competencies, and norms and identities. However, the focus of this typology is very much on the analysis of possible effects of transnational activities, lifestyles and attitudes (for example on gender relations), and less on the differentiation of specific transnational traits and practices on an individual level.

Another well-known systematisation of transnational activities is the one developed and applied in the context of the “Comparative Immigrant Entrepreneurship Project” (CIEP). Portes et al. (1999, 221-22)12 provide a summary of this approach which differentiates between economic, political, civil-societal, and cultural transnational activities. In their comments on the classification, Itzigsohn and co-authors (1999, 324) stress that the division between these groups of activities is sometimes not as straightforward as one would like. This is due to the circumstance that a particular activity might be seen as belonging to one category or another, depending on the point of view taken by the respective researcher. The authors give the example of fundraising for a political party of the CoO which can be classified as a political activity or an economic one. Consequently, Itzigsohn et al. maintain that their systematisation serves first and foremost as a means of providing clarity to the analysis of a large number of transnationalism indicators. Other authors employ very similar classifications; such as Snel et al. (2006, 292), who differentiate between economic activities (of personal or professional nature), political activities, and sociocultural activities (either in the CoO or the CoR). Interestingly the aforementioned approaches are also similar in that they do neither consider – explicitly – the role of social networks covering more than two countries, nor of knowledge of languages other than those of the CoO and CoR. This may be partly due to the fact that much of the empirical and theoretical literature on migrant transnationalism concentrates on enduring transnational relations between migrants in a specific CoR and their co-nationals in their CoO. However, migrants are at least as likely to have personal contacts in third countries as non-migrants. While not limited to this group, these could, for example, be individuals from the migrants’ home region who migrated to third countries.

12 This typology has consequently been applied in a number of influential publications which build on the CIEP dataset. See Schunck (2014, 48-50) for a short discussion and the project archive (https://www.princeton.edu/cmd/data/ciep/) for a list of relevant publications.

Theoretical considerations

Broadening the focus to include other aspects besides transnational activities, Theresa Kuhn (2011) differentiates conceptually between the dimensions of ‘transnational background’, ‘transnational practices’ and ‘transnational human capital’. She describes them as follows:

“[T]ransnational background refers to personal traits such as dual citizenship, the fact of being born abroad or belonging to an ethnic minority. […] Transnational practices entail the active, voluntary engagement in contact with non-national actors. It includes stays abroad, relations and regular contacts with non-nationals or people abroad, as well as crossing borders for work, shopping or leisure. Finally, transnational human capital refers to knowledge and skills that enable people to perform in a transnational environment.” (Kuhn 2011, 814, italics in the original)

While this classification is very helpful – as will be shown below – it also needs further explanation. If examined only superficially, it could be argued that this classification fuels the criticism of transnationalism as just another catch-all concept; at least as far as migration research is concerned. This stems not only from the above definition of transnational background but also from its operationalisation in Kuhn’s research, which is based on respondents’ foreign nationality and their parents’ country of birth (Kuhn 2011, 820). Thus, if ethnic origin and migration experience are considered sufficient to define a person as transnational, all migrants and their immediate descendants would have to be regarded as such. It should be noted that Kuhn did not develop her model in the context of migration research, but to study the support of European integration amongst – mostly – sedentary populations. Furthermore, it can be argued that she is not proposing the above-described simplification. Her argument is rather that an individual with familial relations in another country or with foreign language skills (i.e., transnational human capital) is more likely to be involved in transnational activities than a person who lacks both. In this sense, the differentiation should be made between transnationalism as a theoretical construct and an analytical tool, on the one hand, and the empirical degree of transnationalisation of an individual’s daily life, on the other hand. It is the latter which can be captured by Kuhn’s third dimension: transnational practices. More specifically transnational background, cross-border activities or foreign language skills alone do not automatically lead to a transnational life. However, a person who engages in numerous transnational activities will most likely acquire transnational human capital and, in turn, these combined features will increase the probability of the consolidation of a transnational lifestyle. Generally speaking, Kuhn’s approach promises to be of great help allowing a differentiation between three categories that characterise an individual’s degree of transnationalism. Therefore, the present work will apply a

Theoretical considerations multidimensional understanding of the phenomenon which incorporates the main aspects of Kuhn’s model and differentiations that are closely related to those commonly used in migration research.

The fact that this work will analyse transnationalism in the context of migrant and not mostly sedentary populations leads to a significant modification of the dimension of transnational background. In her work Kuhn operationalises this feature as being born in another country than the CoR or having parents who were born outside the CoR (Kuhn 2011, 819; Kuhn 2015, 33). Therefore, this understanding of transnational background first builds on an individual’s formal belonging to more than one state, and second on his or her membership in certain social groups based on ancestry. In the context of Kuhn’s research, this operationalisation appears appropriate as her data include both individuals with and without migration experience. For the present analysis, however, an unmodified adoption of this approach would not be productive, given that it would attribute the same status to all respondents. Nevertheless, the reference to parents and the belonging to a minority group indicate that this dimension is to a substantial degree linked to the social origin and social relations of individuals, and also to opportunity structures and obligations which derive from them. The important role of personal contacts across borders in the establishment and maintenance of individual transnationalism has also been stressed by various other authors (e.g., Glick Schiller, Basch, and Szanton Blanc 1999, 85-88; Portes, Guarnizo, and Landolt 1999, 220; Mau 2007, 94-95). Therefore, the analysis of transnational background will concentrate on cross-border connections, i.e., transnational social networks, as one of its most important aspects which will allow distinguishing different levels of individuals’ embeddedness in transnational relations.

Besides the existence of transnational networks, this work will pay detailed attention to their geographic dispersion in order to go beyond the binational analytical approach often taken in empirical studies on migrant transnationalism. Guarnizo and Smith, in fact, stress the importance of the geographic dimension, writing:

“Transnational practices, while connecting collectivities located in more than one national territory, are embodied in specific social relations established between specific people, situated in unequivocal localities, at historically determined times.” (Guarnizo and Smith 1998, 11)

Theoretical considerations

Elaborating on their point, the authors argue that the countries which individuals migrate to would provide different opportunity structures and restrictions which, in turn, would influence whether and how migrants engaged in transnational activities. Hence, by providing a structural and social framework upon which interactions could unfold, nation states affect the geographic dimension of transnational spaces from a macro perspective. At the same time, the above-cited argument also hints at the importance of individual connections on a micro level, i.e. to social relations which unfold between individuals and endure over long distances. From this perspective, it is not just the location of the migrants but also of their contacts, that shape the extent and the degree of pluri-locality of the respective transnational network. The relevance of interpersonal networks is also stressed by Mau (2007, 94-95), who points out that the flows of goods, ideas, and information establishes social relationships between individuals while pre- existing social ties simultaneously enable the continuance of such flows. Hence, the geographic extent of a persons’ social network can be seen as an indicator of the geographical space in which transnational activities (e.g., regular communication, money transfers, or the sending and receiving of goods) might occur. Consequently, the present work will examine the geographic scope of transnational social networks reported by Turkish and Romanian migrants residing in the three countries under observation.

In contrast to transnational activities and networks, the concept of transnational human capital is less straightforward. Therefore, it is first of all helpful to examine how the concept is understood in the literature. Building on earlier work by such authors as Gary Becker and Pierre Bourdieu, Gerhards and Hans provide the following definition:

“Human capital […] consists of the person-related knowledge and skills of people which allow them to be successful in different social fields. Transnational human capital refers to those stocks of knowledge and personal skills that enable a person to operate in different fields beyond the individual nation-state.” (Gerhards and Hans 2013, 100, italics in the original)

This definition underlines the two fundamental principles of transnational human capital: first, it is bound to a particular individual, and second, it eases interactions in social and cultural realms into which the respective individual was not born and first socialised. The latter observation might also explain why this dimension of transnationalism is rarely discussed within migration research. As argued above, transnationalism is, on the one hand, conceptualised as a phenomenon that potentially connects people, organisations, and institutions in a variety of places. However,

Theoretical considerations on the other hand, it has mostly been empirically analysed as a phenomenon linking only two geographic spaces (i.e., the CoO and the CoR of migrants). If the ‘different fields’ in which the human capital allows to operate in are understood as being situated in the CoR, the respective knowledge and skills are more logically discussed in the context of integration than of transnationalism. In this case, such human capital would consist, for example, of CoR language knowledge and a familiarity with cultural norms valid in this country. However, in this study transnationalism is not understood as linking only two but potentially a much higher number and diversity of social spaces. Thus, human capital which is in its utility not limited to either the CoO or CoR has to be seen as a very relevant aspect of migrant transnationalism in this context.

In line with the call for analytical clarity cited above, the third specified dimension – transnational practices – will be supplemented with a subdivision mainly inspired by the approaches of Portes et al. (1999), Itzigsohn et al. (1999) and Snel et al. (2006). This follows Schunck’s recommendation, who, based on his empirical findings, calls for “a more precise distinction between different types of transnational activities” (2014, 251). He does so arguing that the influence of various indicators, such as income and labour market integration, may vary in the nature and direction of their relation with different types of transnational practices. Hence, this analysis will differentiate, within the dimension of transnational practices, between communication, mobility, consumption, economic activities, and political activities. Further details on the operationalisation of the discussed dimensions and subdimensions of transnationalism are provided in chapter 3.4.

2.4 Transnationalism and integration – Mutually exclusive or coexisting? As previously mentioned, the possible relation between transnational activities or orientations of migrants and their integration in the CoR is not only a relevant topic in academia but also prominent in public and political discussions in many receiving countries. One of the primary goals of this study is to clarify the nature of this relationship between the two above-described social processes.

In public and political discussions on this topic, it is, however, usually not the term transnationalism itself which is used but a broader description of the same phenomenon such as ‘migrants’ continuing ties to their homeland’. Regarding public debates, authors such as Portes (1999, 469), Levitt (2002, 178), Snel et al. (2006, 287), and Mügge (2016, 110) state on a general

Theoretical considerations level that large parts of the population in immigration countries would often view enduring relations of migrants to their CoO as an obstacle to their integration. Going one step further, Waldinger (2010) and Bouras (2013) empirically show the validity of this general assessment of public opinion in the United States, France, and the Netherlands. They do so using data from the International Social Survey Program (ISSP) and media analysis, respectively. In addition, research on topics like dual nationality and absentee voting indicates that the persistence of transnational practices and the orientation of migrants towards more than one society are often perceived or portrayed as signs of conflicting loyalties (e.g., Bauböck 2005; Aleinikoff and Klusmeyer 2001; Escobar 2007; Jones-Correa 2001). An academic who is often referred to in public discourse on migration and integration related issues is the late Samuel Huntington. Huntington’s work certainly depicts transnational activities and the social incorporation of immigrants in the CoR as fundamentally antagonistic. In his book “Who are we?” Huntington (2004) described transnationalism, fostered by parts of the immigrant population and by members of the political, economic, and intellectual elites in the United States, as undermining the alleged foundations of US-society and as having negative economic implications for its population. In his arguments, it becomes apparent that he understands the relation between integration and transnationalism as a zero-sum game. He applies this logic not only to complex political loyalties (Huntington 2004, 271) but also to supposedly clear-cut economic transactions arguing that every Dollar sent by a migrant to his or her CoO is a Dollar less spend in the United States and therefore a loss to its economy (Huntington 2004, 272). However, as appealing as this black-and- white scenario might be to parts of the broader public, it oversimplifies reality. This view does, for example, not account for the possibility that remittances increase the spending capacity of households in the migrants CoO which might, in turn, have a positive influence on CoR exports. Nor does it consider that transnational economic activities have been shown to have a positive effect on immigrants’ income (Portes, Guarnizo, and Haller 2002, 293) which, in turn, increases their spending capacity in the United States and might allow them to invest in housing, education or to participate in social activities.

Researchers engaged in the investigation of migrant transnationalism likewise conceptualised transnationalism in the earlier stages of its discussion as an alternative to the model of immigrant integration or assimilation in the CoR (e.g., Glick Schiller, Basch, and Blanc-Szaton 1992b, 16; Portes, Guarnizo, and Landolt 1999, 228-29). However, as the theoretical discussion advanced

Theoretical considerations and studies evaluating empirical evidence of migrant transnationalism grew in number this position has largely been replaced by assessments which increasingly understand integration and transnationalism as phenomena which are not conflicting but possibly even interconnected.13

Various authors indicated, however, that systematic quantitative research on the extent to which migrants engage in transnational activities in general and on the relationship between transnationalism and integration, in particular, has been scarce until recently (Snel, Engbersen, and Leerkes 2006; Faist, Fauser, and Reisenauer 2013; Schunck 2014). The first major quantitative study which focussed on this topic was the previously mentioned “Comparative Immigrant Entrepreneurship Project” (CIEP) which surveyed 1,202 migrants from Colombia, the Dominican Republic and El Salvador in four American cities14 in late 1997 and early 1998 (Itzigsohn and Giorguli Saucedo 2002, 773; CMD 2017). As the title suggests, this project concentrated on migrants’ professional and commercial activities in a transnational context. However, it also collected data on aspects of cross-border political engagement and sociocultural activities. Consequently, the scholars involved in this research were able to produce a number of highly cited articles analysing various aspects of the interplay between transnationalism and migrant integration (e.g., Itzigsohn and Giorguli Saucedo 2002; Portes, Guarnizo, and Haller 2002; Guarnizo, Portes, and Haller 2003). Further analysis of migrant transnationalism in the United States based on a different dataset was carried out by Waldinger (2008). During the last decade, some studies in this field have also been carried out in two other ‘classic’ immigration countries, namely Australia (O'Flaherty, Skrbis, and Tranter 2007) and Canada (Ley 2013). Starting with an article by Snel, Engbersen and Leerkes (2006), the relationship between integration and transnationalism has become a topic of quantitative investigation in Europe, too. This article uses data collected on immigrants from six different countries residing in the Netherlands. Later quantitative work by Schans (2009), Dekker and Siegel (2013), and Bilgili (2014a; 2014b)15 also focuses on the Netherlands. Furthermore, Carling and Hoelscher (2013) have published findings on the relation between integration and remitting behaviour of migrants in Norway. Finally, building on data of the ‘Socio-Economic Panel’ (SOEP), Schunck (2014) produced enlightening findings on the topic with regard to immigrants in Germany.

13 See: Itzigsohn and Giorguli Saucedo 2002, 770; Morawska 2002, 161-62; Levitt 2002, 192; Levitt and Glick Schiller 2004, 1003; Portes, Escobar, and Arana 2009, 121; Tsuda 2012, 646; Faist, Fauser, and Reisenauer 2013, 100. 14 New York City, Los Angeles, Washington, DC, and Providence. 15 The studies of Dekker and Siegel (2013) and Bilgili (2014a; 2014b) build on data from the same research project.

Theoretical considerations

Besides the grouping by CoR, a relevant way of categorising these studies is to review which aspects of migrant integration and transnationalism they focus upon. After applying the discussed four-dimensional framework put forward by Esser (see chapter 2.2.2) a certain pattern becomes visible.

All of the 12 publications listed here include in their analysis the relation between structural integration and transnationalism. In most cases, this aspect is, however, not the prime focus of the respective analysis. Instead, indicators of structural integration, such as labour market status and income are included as control variables alongside education. Second most often discussed are elements of cultural integration, such as knowledge or use of CoR language (see for example Schunck 2014). In addition to this aspect Bilgili (2014a), for instance, includes an indicator measuring the consumption of CoO media and art.16

Interestingly social integration is explicitly discussed in only three of the studies which use data from the Netherlands. While Snel et al. (2006) do so by determining the number of native Dutch in respondents’ social networks, Bilgili (2014a) and Dekker and Siegel (2013) measure how often respondents spent time with natives in a private setting. Admittedly, two of the CIEP studies (Portes, Guarnizo, and Haller 2002; Guarnizo, Portes, and Haller 2003) also report indicators of the size and scope of respondents’ social networks. However, the authors do not differentiate whether the respective networks consist primarily of co-ethnics, other migrants, or natives of the United States.

A similar, yet in terms of CoRs somewhat more diverse picture presents the case of identificational integration. In the data of Carling and Hoelscher (2013, 945), respondents in Norway had to evaluate their perceived level of belonging on a 7-point rating scale. This is similar to the measurement reported by Ley (2013, 927), which is based on a Likert scale on which respondents had to indicate how often they felt Canadian. Both studies, however, use indicators of transnationalism as dependent variables and identificational integration as one of several predictors. While Carling and Hoelscher concentrate on remittance behaviour as a single transnational activity, Ley uses separate models to predict determinants of various aspects of transnationalism separately. Consequently, the focus and analytical strategies of both papers do not allow their authors to discuss the relation between identificational integration and broadly

16 The remaining studies which include elements of cultural integration in their analysis are: O'Flaherty, Skrbis, and Tranter (2007), Waldinger (2008), Schans (2009), Carling and Hoelscher (2013), and Dekker and Siegel (2013).

Theoretical considerations measured transnationalism. A more complex measurement of identification is used in the study of Snel et al., whereby respondents were asked to position themselves with regard to a number of topics relative to native Dutch, co-nationals in the CoR, and compatriots in the CoO and in third countries (2006, 290). Finally, Schunck chooses a very different approach and uses the intention to settle permanently in Germany to operationalise the affective aspect of identificational integration when using this dimension as dependent variable (Schunck 2014, 33, 276). When treating the same aspect of integration as independent variable, he uses the acquisition of CoR citizenship as an additional indicator (Schunck 2014, 119-20). However, it is questionable whether naturalisation is an ideal choice to measure the degree of identification with the CoR. On the one hand, the acquisition of citizenship is usually bound to a number of institutional preconditions (e.g., the requirement of a specific minimal period of residence), which are not fully under the individual’s control. At the same time, it is conceivable that a migrant might identify highly with a CoR without fulfilling certain naturalisation criteria. On the other hand, citizenship is tied to a number of advantages, rights, and obligations (e.g., visa waiver programs, voting rights), which might be attractive to residents even if they do not particularly identify with the specific country. In a nutshell, individuals who identify highly with a country might not be able to take up citizenship for formal reasons, whereas others might become naturalised for purely pragmatic reasons while lacking a particular feeling of belonging to the country in question. Consequently, it seems more appropriate to consider naturalisation an aspect of structural integration, which would also be more consistent with Esser’s arguments (2000, 272). Schunck is indeed aware of the ambiguities associated with the nationality indicator and briefly mentions some of the above-stated arguments himself (Schunck 2014, 119-21).

In conclusion, it can be said that the least information is available regarding the relation between aspects of social and identificational integration, on the one hand, and transnationalism on the other. Consequently, this work will focus on these dimensions. Aspects of structural and cultural integration will nevertheless be reviewed. However, they will not take a central position in the discussion and be treated as independent rather than dependent variables. The decision to investigate social integration is also grounded in the general importance of social interactions. This significance is stressed by Esser (2000, 273-74; 2001, 11) who argues that social interactions can support the acquisition of cultural and social capital which can further strengthen the social position and life possibilities of migrants in the CoR. Moreover, the additional focus on

Theoretical considerations identificational integration, promises to be a fruitful contribution to the field. This is partly due to the fact that few earlier studies included this aspect. More importantly, however, this analysis will not only take the migrants’ identification with the CoR into consideration but also their identification with the town they live in. This decision is first based on the theoretical considerations of Lawler (1992). He argued that in the case of nested collectivities17 individuals would most likely identify strongest with the group or entity which they perceived to be most influential with regard to their opportunities in life, the support it offers, and the degree to which it can be influenced.18 Consequently, the closer and less abstract local entity should be more relevant for migrants who do not possess CoR nationality. Second, in addition to these theoretical considerations, empirical results suggest the existence of a stronger identification with the town of residence (ToR) than with the CoR (Crul and Schneider 2010, 1262). If the assumption holds true that the level of identification differs between the smaller (ToR) and the bigger (CoR) unit it will be of interest if the (transnational) factors influencing these identifications (and their effects) will differ as well.

In addition to the focus on different dimensions of integration in existing studies, cross-country comparisons of their findings are not only hampered by methodological differences but also by the great variety of transnationalism indicators included in the analyses. Frequently considered aspects include different forms of monetary remittances19 and visits to the CoO20. In general, the number of transnationalism features which authors take into view varies greatly. While some concentrate on one or two specific aspects (e.g., O'Flaherty, Skrbis, and Tranter 2007; Carling and Hoelscher 2013; Schunck 2014), others take more facets of transnationalism into view. However, Snel et al. (2006) who report findings on a total of 17 indicators are certainly the exception. Hence, the existing literature shows a high degree of diversity not only with respect to sample sizes and target populations but also regarding the included dependent and independent variables.

17 An example would be cities which are nested in regions which are nested in states. 18 This condensed argument mostly builds on the more detailed thoughts put forward by Lawler (1992) in his second (p. 331), third (p. 333) and fourth proposition (pp. 334-5). 19 Snel, Engbersen, and Leerkes 2006; Waldinger 2008; Schans 2009; Carling and Hoelscher 2013; Dekker and Siegel 2013; Ley 2013; Itzigsohn and Giorguli Saucedo 2002; Schunck 2014. 20 Itzigsohn and Giorguli Saucedo 2002; Portes, Guarnizo, and Haller 2002; Guarnizo, Portes, and Haller 2003; Snel, Engbersen, and Leerkes 2006; O'Flaherty, Skrbis, and Tranter 2007; Waldinger 2008; Dekker and Siegel 2013; Ley 2013; Bilgili 2014a; Schunck 2014.

Theoretical considerations

Most of the above-mentioned studies come to the conclusion that there is no evidence of a general negative relation between migrants’ transnational engagement and their integration in the CoR.21 In order to reduce repetitiveness throughout the text, the following section will concentrate on a number of specific studies which found considerable negative associations between some indicators of transnationalism and integration (often whilst finding positive relationships between other features of both phenomena). This will serve to illustrate the complex nature of both aspects. A broader set of findings will be reported in the last part of this chapter in which a number of hypotheses will be presented.

As mentioned, various possible determinants of transnational engagement are discussed in the literature. In particular, aspects of structural integration are often a key focus as they are vital to the social mobility of both migrants and CoR population. The two indicators most frequently discussed in this context are (perceived) discrimination and income levels. Both are likely to be related since actual discrimination might hinder qualified migrants from receiving well-paid employment, while individuals who are not self-aware of significant gaps in their qualifications might attribute their possible failing on the labour market to alleged discrimination. At the same time, individuals who hold comparatively well-paid positions might be more aware of individual or institutional discrimination due to their professional experience in these positions and their educational background (Schunck 2014, 270; Snel, 't Hart, and Bochove 2016, 515). However, migrants with higher income may also be more likely to engage in certain types of transnational activities which require the use of monetary resources (e.g., travel or regular remittances). As previously mentioned, Itzigsohn and Giorguli Saucedo (2002) coined the terms reactive and resource dependent transnationalism for these mechanisms, which they differentiate from a third type, called linear transnationalism. While stating that transnationalism and integration do generally not contradict each other, the authors argue that the relation between transnational activities and sociodemographic variables is complex and may vary depending on the nature of the observed activity. Nevertheless, they find some evidence for all three mechanisms. This is exemplified by the fact that discrimination and a negative attitude towards the CoR show positive associations with the index of sociocultural transnationalism used by the authors, providing support for the reactive transnationalism argument. Furthermore, being unemployed as opposed to being an employee or entrepreneur is negatively related to transnational engagement, which

21 This applies to: Portes, Guarnizo, and Haller 2002; Guarnizo, Portes, and Haller 2003; Snel, Engbersen, and Leerkes 2006; Schans 2009; Carling and Hoelscher 2013; Dekker and Siegel 2013; Schunck 2014.

Theoretical considerations corresponds to the resource dependent transnationalism hypothesis. Finally, the authors see the positive relation between transnationalism and a stated wish to return to the CoO at some point, as supporting the idea of linear transnationalism (Itzigsohn and Giorguli Saucedo 2002, 784-85). Hence, on the one side, the fact that no negative relations are found between transnationalism and labour market integration, CoR citizenship, or the time spent in the CoR underlines that there is no proof of a general negative relation between transnationalism and integration. On the other side, the existence of reactive transnationalism indicates that the combination of a lack of opportunities with a lack of in interest in integration can be associated with a higher level of transnationalism. However, the direction of the causal effect is not obvious in this case. Following the approach of Itzigsohn and Giorguli Saucedo, Snel et al. (2016) recently also found evidence of reactive transnationalism in a study of middle-class migrants in Rotterdam. As in the before mentioned case, perceived discrimination negatively correlated with different types of transnational activities in their data. However, the findings of this study are derived from a comparatively small sample and should be interpreted with care.22

Two contributions that present significant negative relations between aspects of integration and transnationalism as part of their main findings are those of Ley (2013) and O’Flaherty et al. (2007). The former finds negative relations between indicators of sociocultural integration and transnationalism amongst predominantly Asian migrants in Vancouver.23 He operationalises this specific dimension of integration using the acquisition of Canadian citizenship, a sentiment of feeling Canadian, and a variable indicating in which language the respondents completed the survey. However, according to the systematisation of integration used in the present analysis, all three variables are considered indicators of different dimensions of integration. Namely, structural, identificational, and cultural integration, even though it could generally be questioned whether the survey language variable is an appropriate indicator of integration. Nevertheless, Ley’s results show that those respondents who acquire Canadian citizenship, who tend to feel Canadian more often, and who answered the survey questions in English instead of their native language are significantly less likely to engage in transnational activities. The author himself stresses that these results diverge from earlier findings of Portes et al. (2002) and Snel et al.

22 The sample of this study consisted of 225 individuals of whom only 140 (i.e., 62 percent) were migrants. The remaining 84 participants were children of migrants but did not have migration experience themselves (Snel, 't Hart, and Bochove 2016, 530). The authors control for this aspect in their analyses. 23 For a detailed description of the ethnical composition of the sample analysed by Ley see Hiebert and Ley (2006, 83).

Theoretical considerations

(2006). Whereas Ley takes a number of transnational activities into view, such as remittances, and property in or travels to the CoO, O’Flaherty and his co-authors concentrate on investigating the latter aspect with regard to immigrants in Australia. In their analysis, they also find evidence of a negative relation between cultural assimilation and transnationalism. However, again it is instructive to closely examine the author’s operationalisation of cultural assimilation, which consists of the general life satisfaction, the acquisition of Australian citizenship (respectively the desire to acquire it), and the level of knowledge of the English language. Hence, if again the understanding of the present work is applied, only one indicator – namely the command of English – falls into the dimension of cultural integration. Interestingly, this is also the one aspect amongst these three for which the authors actually report a positive association with transnational mobility.24 In line with Ley’s findings the indicators for which the authors find negative effects are at the intersection between structural and identificational integration.

The last example of a negative relation stems from the study of Bilgili (2014a). On a general level, the author shares the assessment that integration and transnationalism are non-conflicting features. Bilgili even reports a positive relation between membership in civic organisations in the CoO and CoR, as well as between the consumption of media and art originating in the CoR and the CoO. However, her findings also indicate that respondents who only spoke the language of their CoO at home showed a significantly higher degree of transnationalism across different indicators (contact with family and friends in the CoO; visits to the CoO; membership in CoO associations and consumption of CoO art and media) than those who spoke mostly or only Dutch. It follows that, if the use of CoR language at home would be understood as an indicator of cultural integration, this result would mean that such integration would be negatively associated with transnational engagement. However, as previously mentioned, the consumption of CoR media and art as another indicator of cultural integration is positively related to the corresponding transnational orientation (i.e., the consumption of CoO media and art). Hence, in this case, a positive relation between cultural integration and transnationalism is apparent. Unfortunately, the author did not include the indicator of CoR media and art consumption in her analysis of the determinants of the remaining transnational activities she focused on. Therefore, the relation between a more complex operationalisation of cultural integration and a broader understanding of transnationalism is not addressed in her analysis.

24 Besides this, additional positive associations are found by the authors for example between transnationalism and home ownership in the CoR.

Theoretical considerations

These examples show that quantitative findings regarding the relation between transnationalism and integration are partly inconclusive or even conflicting. David Ley (2013, 935) argues that differences in the cultural and ethnical composition of the studied target populations, different context factors in their CoR, and the diversity of possible relations between CoO and CoR might partly explain these diverse findings. He justifies this assumption not only by the fact that his results diverge from those presented by Portes et al. (2002) and Snel et al. (2006) but also points out that his sample largely consists of migrants from East, South and Southeast Asia instead of respondents of Latin-American, respectively North-African or Turkish descent. His assumption is also supported by several studies showing relevant differences in transnational engagement between migrants from different CoOs (e.g., Guarnizo, Portes, and Haller 2003; Snel, Engbersen, and Leerkes 2006; Schans 2009).

These issues point directly to the relevance of comparative research in this area. All of the above- mentioned studies included more than one migrant group, and some form of comparison by including group identifiers and reporting differences between migrants with varying backgrounds. Consequently, these works implicitly acknowledge Itzigsohn’s and Giorguli Saucedo’s assessment that “differences in the determinants of transnational participation can only be established through comparative research.” (2002, 767). Two types of comparison could be helpful in identifying such differences. First, varying outcomes may be observed if migrants from different CoOs are surveyed in the same CoR. Second, differences may be observed if the national background of the respondents (i.e., the CoO) is kept constant, while various CoRs are taken into view. However, a limitation of the cited studies consists in the fact, that all rely on data which had been collected in one CoR only. Hence, they can only address the first type of comparison. However, it should be noted that an additional quantitative study, authored by Castañeda, Morales, and Ochoa (2014), does partially address this gap.25

The mentioned article (Castañeda, Morales, and Ochoa 2014) builds on data collected in two cities in the United States (New York and El Paso) and in the French capital Paris. While this study is very interesting and offers some valuable insights, it also suffers from some methodological weaknesses. These are mostly related to the fact that the authors used data from

25 Another study (Braun and Müller 2012) does indeed include migrants from various EU countries residing in other EU countries. However, the authors’ analysis concentrates on the determinants of European identification in comparison to CoR and CoO identification and, therefore, has a somewhat different focus than the present work and the other above-discussed contributions.

Theoretical considerations three different surveys which were not carried out simultaneously and which realised very different sample sizes. More precisely, the participants were 1,038 Latinos in El Paso and 363 in New York, and 65 respondents of Northern African origin in Paris. Furthermore, these samples included descendants of migrants without own migration experience, which has the consequence that the French sample only contained 35 migrants in the literal meaning of the term (Castañeda, Morales, and Ochoa 2014, 312, 320). The cross-national comparability of these data is also hindered by the fact that they stem from migrant residents in the two mentioned CoRs who did not originate in the same countries or similar regions. Consequently, observed differences can only to a limited extent be attributed to context effects in the two CoRs since the populations under investigation differ in respect to several key variables directly or indirectly related to their CoO (e.g., first language, cultural background, social structure of the CoO, possible impact of the educational and political systems of the CoO on individuals). Additionally, a composite variable was used to measure socio-economic status, which is one of the key predictors in this analysis. Unfortunately, the authors constructed this index from different variables in El Paso than in New York and Paris (Castañeda, Morales, and Ochoa 2014, 313). Finally, the quantitative analysis presented by the authors is mostly based on descriptions of bivariate relations. Due to these methodological limitations, the results of this study should be interpreted with caution. As a consequence, the use of this study as a point of reference in the present text will be limited to parts of the descriptive analysis.

Thus, in agreement with Schunck (2014, 291-92) it can be concluded that the lack of cross- national quantitative studies investigating the relation between transnationalism and integration in a comparative manner has still not been sufficiently addressed.

Theoretical considerations

2.5 Summary and hypotheses In conclusion, migrant integration is understood in this work as a multidimensional process in which cultural, structural, social, and identificational integration can be differentiated. This process is not necessarily linear and can change its velocity and even its direction. Any individual with or without migration experience can, for example, be fully incorporated in a specific social space at one point in time and not participate in it at a later one. This could either be due to changes in the individual or to changes in the social space to which the individual did belong. This also hints at the fact that integration is a two-way process. As such, it is influenced not only by the traits of the immigrant but also by the opportunities and obstacles present in the CoR. In the discussion of relations between transnationalism and integration, this work will concentrate on the dimensions of social and identificational integration.

Transnationalism is understood as a complex and multidimensional phenomenon too. It consists of individual transnational background and networks, transnational human capital, and transnational practices. Hence, it is not only comprised of the individuals’ enduring cross-border practices but also of their social connections and the skills which allow them to interact successfully in social contexts that connect people and organisations in more than one country. Whereas the focus of current empirical literature on migrant transnationalism is characterised by a CoR-CoO dichotomy, this work will expand this view by examining the geographic scope of transnational social connections and activities. This study will, moreover, take a cross-national comparative approach in order to address the respective gap in the current research. A key question in this analysis is whether and how inclusion and participation in one context (i.e., a transnational space), is related to inclusion and participation in another (i.e., the CoR). The following presents hypotheses which will, on the one hand, consider factors that might determine transnational practices and, on the other hand, account for the relation between transnationalism and integration.

A first aspect which can be assumed to be related to the degree to which individuals engage in transnational activities is gender. The literature, however, shows mixed results in this regard. Several studies have found that men are more likely than women to engage in political transnationalism (Guarnizo, Portes, and Haller 2003, 1235; Waldinger 2008, 15; Dekker and

Theoretical considerations

Siegel 2013, 12) and to remit money (Waldinger 2008, 15; Carling and Hoelscher 2013, 950). Whereas Itzigsohn and Giorguli Saucedo found the same tendency regarding public sociocultural transnationalism (2002, 787), other results indicated that women are more likely to engage in social transnational practices, such as transnational communication (Schans 2009, 1175). However, these findings are only at first glance at odds with each other as Itzigsohn and Giorguli Saucedo emphasise that their operationalisation of sociocultural transnationalism pertained to the public sphere and that the role of gender might be different regarding transnational practices in private settings (2002, 790). Several of the cited studies found gender effects for one type of transnational activity but not for another. For example, Schans (2009) did not find any significant effects regarding the sending of remittances, and Dekker and Siegel (2013) did not report significant effects for economic and sociocultural transnationalism. Finally, Waldinger (2008, 15) did not find any gender effect on the probability of visits to the CoO. Unfortunately, Schunck (2014) did not report any results regarding gender, which he used as a control variable. Nevertheless, the findings reported by the other mentioned studies would suggest that some gender differences in transnational engagement do indeed exist. Regarding the nature of these differences, the hypothesis is that men are more likely to engage in comparatively formalised transnational political activities, such as absentee voting (HT1.1), while women are more likely to engage in transnationalism on an informal and private level, e.g. transnational communication (HT1.2). Whether and how gender may be related to the other types of transnational practices cannot be predicted from existing results which are either inconclusive or show no significant effects at all.

The age at which respondents first settle in the CoR could also be expected to influence transnational involvement. Migration at an older age means, for example, that a larger part of the individual’s socialisation took part in a different context than the one this person is currently living in. In theory, as mentioned by Schunck, this could lead to a higher inclination to stay in contact and interact with people in these contexts after migration (Schunck 2014, 118). However, the empirical findings regarding this indicator are ambiguous. In his own analysis, Schunck does not find any significant effects of age at migration. While Dekker and Siegel (2013, 12) report a positive effect of age at migration on political transnationalism of immigrants in the Netherlands, they do not find any significant effects on transnational activities of an economic or sociocultural nature. Waldinger (2008, 15) reports that respondents who settled after the age of 12 were more

Theoretical considerations likely to remit than those who migrated as younger children. However, his data show no significant associations between this feature and absentee voting or visits to the CoO. Finally, Schans (2009, 1179) reports contradicting findings for different transnational practices. She finds that age at migration is positively associated with remitting money, but negatively related to the frequency of contact with people in the CoO. Due to the ambiguity of these results, no clear hypothesis can be formulated with respect to the age at which respondents migrate and their engagement in transnational practices.

A second time-related indicator that is often included in analyses of transnational engagement is the duration of residence in the CoR. In this context, it is important to keep in mind that most empirical analyses focus on activities which are directed towards the CoO. In these cases, it is mostly assumed that the duration of residence is negatively related to the engagement in transnational activities, as individuals increasingly orient their lives towards the CoR and as the geographic distance results in an increasing social distance to individuals in the CoO over time. Indeed, empirical studies find a negative relation between the duration spent in the CoR and the number of visits to the CoO (Schunck 2014, 222-26), frequency of communication with family members abroad (Schans 2009, 1176), and participation in CoO elections (Waldinger 2008, 15, 18). In contrast, Guarnizo et al. (2003, 1238) find a positive effect of the time spent in the CoR on political transnationalism. However, these authors use a composite measure which does not include absentee voting but instead consists of six specific features and activities (e.g., membership in a political party of the CoO and remitting money for community projects in the CoO). As will be shown in the next chapter, their measure of political transnationalism differs substantially from the one used in the present analysis. Using the same data as the aforementioned authors, Portes, Guarnizo and Haller find a positive effect of the duration of stay on transnational entrepreneurship (2002, 290). However, this is a very specific and in many regards (i.e., logistically and legally) complex type of transnational activity, which only a comparatively small number of migrants will engage in. In this case, the positive effect of the duration of stay could, for example, be related to a possible accumulation of financial resources over time, which are necessary to start an enterprise. Unfortunately, the data used here do not allow for the investigation of migrant entrepreneurship. On a more general level, empirical studies often focus on monetary remittances as an indicator of economic transnationalism. However, they usually do not find any significant correlation with the duration of stay in the CoR

Theoretical considerations

(Schans 2009, 1173; Carling and Hoelscher 2013, 950; Dekker and Siegel 2013, 12). Nevertheless, given the above-mentioned results, the hypothesis is that the longer their stay in the CoR the less migrants are likely to engage in transnational practices (HT2).

Certainly, the age of respondents at the time of the interview is an important control variable which is routinely included in analytical models in the social sciences. However, age is also the sum of age at migration and duration of stay in the CoR. This has the effect that only two of these three variables can be included in regression analysis in order to avoid collinearity. Since the effect of the age variable is assumed to be less important than age at migration and duration of stay regarding transnational involvement and, especially, regarding integration, age itself is not included in the regressions.

For their analysis of determinants of transnationalism Snel et al. (2006, 294-95) built a cumulative index of transnational activities covering a wide range of economic, political, and sociocultural activities. In using this indicator as dependent variable, they do not find any significant effects of education or labour market integration. However, other analyses which have looked at individual indicators or a smaller number of transnational activities, have shown that higher formal education is often positively related to the engagement in transnational activities. Such results were presented with respect to social (Schans 2009, 1176), sociocultural (Itzigsohn and Giorguli Saucedo 2002, 783), and political transnationalism (Guarnizo, Portes, and Haller 2003, 1230; Dekker and Siegel 2013, 12). Hence, it is possible that opposing effects of educational levels with respect to different indicators included in the index of Snel et al. (2006) cancelled each other out in the ex post created measurement. The present analysis will use several indicators of transnationalism differentiating between various forms of cross-border engagement, thus resembling the approach of the second group of studies. Consequently, respondents with a higher level of formal education are expected to engage more often in transnational activities than those with lower education (HT3).

Regarding labour market integration, two different assumptions could be made: On the one hand, individuals who participate in the regular labour market are likely to have more financial resources at their disposal which could be used and invested transnationally. This would correspond to the resource dependent transnationalism hypothesis formulated by Itzigsohn and Giorguli Saucedo (2002). On the other hand, these individuals might also have less time to

Theoretical considerations engage in transnational activities than those who are not participating in the labour market. Following the latter logic, Schunck (2014, 122) expected to find a negative relation between integration in the labour market and transnationalism. However, those cross-border activities that consume more time are also those for which financial capital is imperative, such as travelling and transnational investments, suggesting that they might indeed rather be positively related to labour market integration. In accordance with this assessment, several authors – with Schunck amongst them – find a positive correlation between labour market integration and the sending of remittances, or with other forms of economic transnationalism (Waldinger 2008, 19; Schans 2009, 1173; Carling and Hoelscher 2013, 950; Dekker and Siegel 2013, 11; Schunck 2014, 250). The findings of Snel and co-authors on immigrant transnationalism in the Netherlands once again also underline the importance of including group indicators as the relationships they found do not pertain to all of the six nationalities surveyed in their study. More specifically, they report a positive effect of labour market integration on professional economic activities for immigrants from the United States, and on social-cultural transnationalism for individuals originally from Iraq (Snel, Engbersen, and Leerkes 2006, 300). Likewise, Itzigsohn and Giorguli Saucedo (2002, 783) find a positive effect of labour market integration on sociocultural transnational activities in the public realm. On the contrary, no significant effects are found in research with regard to the frequency of transnational communication (Schans 2009, 1176), political transnationalism (Dekker and Siegel 2013, 11), and the number of visits to the CoO (Schunck 2014, 244). Regarding the latter aspect, the results of O’Flaherty et al. (2007, 834) are inconclusive, as they find, in different models, indications of positive and negative associations between transnational mobility and fulltime employment. Generally speaking, the majority of findings indicate, however, that economic liberties resulting from the participation in the labour market tend to be more relevant with respect to transnational practices than the time constraints which may result from it. Building on these results, no significant effects of labour market integration are expected with respect to transnational communication, political transnationalism, and transnational consumption. However, a positive effect of labour market integration should be visible with regard to economic transnationalism, given that the participation in the labour market is an indicator of financial stability. Hence, it can be assumed that respondents who are integrated in the formal labour market (i.e., who have a paid job) would be more likely to engage in economic transnationalism than individuals who are not integrated in the labour market (HT4). If the ratio between the investment of time and the necessity of financial security is taken as orientation,

Theoretical considerations transnational mobility would be situated somewhere in the middle. As travelling, on the one side, takes time and, on the other side, requires money it cannot easily be paired with primarily time nor primarily resource dependent practices. O’Flaherty and co-authors (2007, 834) likewise refer to these opposing mechanisms as a likely explanation for their inconclusive results regarding fulltime employment. Adding to this uncertainty regarding possible effects is the circumstance that this analysis, in contrast to Schunck (2014) and O’Flaherty et al. (2007), operationalises transnational mobility not only as visits to the CoO but also includes trips to third countries. Hence, the results can be expected to diverge from earlier findings. Based on these considerations, no hypothesis shall be formulated regarding transnational mobility and its relation to labour market integration.

The logical extension of the above arguments is that there should be a positive relation between the financial situation of the respondents or their households, and the probability of engaging in certain transnational activities. This assumption is supported by empirical findings of Waldinger (2008, 19), Schans (2009, 1173), and Schunck (2014, 250-51), although it does not uniformly apply to all forms of transnationalism. While Dekker and Siegel (2013, 13) did not find a direct significant relation between income and transnational activities, they showed that the experience of financial need is negatively related to political and economic transnationalism.26 Thus, the most direct association is likely with economic transnationalism for which it is hypothesised that the likelihood of respondents to engage in it is higher, the more positive they evaluate their household financial situation (HT5).

Taking the discussed claim of Samuel Huntington into consideration, which asserts that financial resources spent transnationally lead to a drop in investments in the CoR, it is of interest to investigate the relation between the ownership of real estate in the CoR and transnational activities. The former is a type of investment that is, besides the initial costs of acquisition, usually associated with recurring expenses for maintenance, tax-payments and so on. Consequently, Huntington’s claim suggests a negative relation between transnational practices and property ownership in the settlement country. Two studies that investigate remitting behaviour and also include indicators of homeownership in the CoR are those of Marcelli and Lowell (2005) and Menjivar, DaVanzo, Greenwell and Valdez (1998). Both surveys collected

26 In this study, the measurement of financial need was based on an index build from six yes/no questions. In these questions, respondents were, for example, asked whether they could afford adequate heating or eating meat several times a week if they had the desire to do so (Dekker and Siegel 2013, 9).

Theoretical considerations their data on different immigrant groups in Los Angeles County (United States). Contrary to Huntington’s claim, the first study finds a positive relation between homeownership and remitting money to the CoO. The second study does not find a significant association of remitting as such and property ownership in the CoR.27 Marcelli and Lowell propose that a possible explanation for the positive relation may be that migrants who have the financial means to do so, would make somewhat secure investments in the CoR first, before sending money to their CoO for investment or to assist family members (2005, 91). In line with this argument, it could be assumed that real estate ownership signals a certain level of economic stability which might provide individuals with a feeling of security that allows them to invest resources in transnational practices. This idea could also explain the positive relation between homeownership in the CoR and visits to the CoO, as reported by O’Flaherty and co-authors (2007, 834). Taking these findings into consideration, it is hypothesised that respondents who have property in the CoR are more likely to engage in economic transnationalism and transnational mobility than those who do not own such property (HT6). Based on the existing findings, no clear hypotheses can be formulated with regard to other transnational activities. However, it may be that the effects of this indicator correspond to those of the household financial situation and labour market integration.

One of the main goals of this work is the examination of relations between transnationalism and integration. Therefore, possible effects of perceived discrimination have to be taken into account as they are indicative of the degree to which the CoR is experienced as an open society by the individual. According to Itzigsohn and Giorguli Saucedo, discrimination is one of the main factors which could lead to what they describe as reactive transnationalism which emerges as “the result of experiences of discrimination and a negative subjective assessment of the experiences of the immigrant in the country of reception” (Itzigsohn and Giorguli Saucedo 2002, 773). While these authors indeed find that perceived discrimination enhances the probability of engagement in transnational activities, other empirical results are mixed. In agreement are Snel, 't Hart, and Bochove (2016) who find, in their study on immigrants in the Netherlands, a positive effect of perceived discrimination on transnational activities in different domains. However, Schunck only finds a significant effect of discrimination on the amount of money remitted by migrants with no effects on the probability of remitting itself or on the length or number of CoO visits (2014, 223, 248). Carling and Hoelscher (2013, 954) do not find any significant effect of

27 Menjivar et al. (1998, 112) report a negative relation between the remitted amount of money and home ownership. However, this relation is only significant on a 10 percent level.

Theoretical considerations discrimination on the remitting behaviour of immigrants in Norway at all. Consequently, the unknown extent of the relation between discrimination and transnationalism does not allow for the formulation of a hypothesis.

In this research, the general assumption is that transnational background and networks as well as transnational human capital, are positively related to transnational activities. Furthermore, it can be assumed that the effects vary across the different types of transnational activities. Transnational human capital could be especially relevant for dimensions like transnational communication and consumption, which are likely to be associated with knowledge of a third language and country. A positive relation could, for the same reason, also be expected with regard to mobility, since existing knowledge of a country could increase the probability of visiting it again. None of these indicators is, however, usually included in studies of migrant transnationalism. Some results are indeed available regarding migrants’ transnational social networks between the CoR and CoO. For example, Carling and Hoelscher report a significant positive relation between having at least one parent in the CoO and remitting behaviour (Carling and Hoelscher 2013, 950). Possessing a partner in the CoO, however, was not significant in the same model. The general tendency of the findings corresponds nevertheless to the results of Schunck (2014, 248), who found that having parents or children in the CoO highly increased respondents probability to remit money. However, he did not find any effects on the number of visits to the CoO during the 24 months preceding the survey (2014, 227). Whereas no specific hypotheses can be formulated in regard to many types of transnational practices and their relation with the other two dimensions of transnationalism, it can be expected, that the more friends and family members respondents have in the CoO, the higher their probability to engage in economic transnationalism (HT7).

As mentioned above, previous studies repeatedly showed that engagement in transnational activities varied between migrants from different CoOs (e.g., Guarnizo, Portes, and Haller 2003; Snel, Engbersen, and Leerkes 2006; Schans 2009). Consequently, it is to be expected that the group indicators included in this analysis may show significant effects in some cases. However, due to the fact that previous studies focussed on migrants from different CoOs in other CoRs than this research, no specific hypotheses can be formulated for expected deviations between both nationalities within the same CoR or across all three CoRs.

Theoretical considerations

Existing empirical literature suggests that the correlations between transnationalism and integration are limited at best.28 Even though the two dimensions focussed on in this work differ from those reviewed in most studies, as do the included indicators of transnationalism, it nevertheless can be assumed that significant effects are small in number.

As previously argued, the CoR and transnational spaces are understood as two realms in which integration can theoretically occur. However, of the two modes of successful integration which Esser identified as theoretically possible, i.e., multiple inclusion and assimilation, only the latter is according to him empirically relevant while the former is limited to privileged groups, such as children of diplomats (Esser 2001, 21). If transnationalism and integration in the CoR are, in this sense, understood as an either/or constellation, it follows that they should be negatively related to each other. These effects should be strongest between independent variables which directly correspond to the specific indicator of integration. Hence, there should be a significant negative relation between the size of respondents’ social networks amongst CoR nationals in the CoR and the size of their social networks abroad. In other words, the more friends and family a respondent has abroad, the weaker should be their social integration into the CoR (HI1). While identification with the CoO in itself is not a form of transnationalism, this reasoning would nevertheless suggest a similar effect in this regard. Hence, it is hypothesised that the identificational integration in the CoR is weaker, the stronger respondents identify with their CoO (HI2). As the relation between identificational integration on a subnational level and transnationalism has not yet been investigated quantitatively, no clear hypothesis can be made for this aspect.

In their analysis, Snel and co-authors (2006) do not find effects of private or professional transnational economic activities on the number of CoR nationals in their respondents’ networks (Snel, Engbersen, and Leerkes 2006, 302). Likewise, Dekker and Siegel do not find a relation between economic transnationalism and social interactions between migrants and native Dutch citizens (Dekker and Siegel 2013, 11). Furthermore, Schunck does not report significant effects of remitting behaviour on cultural integration (2014, 260). However, Snel and co-authors find that professional economic activities are positively related to the identification with native Dutch. Unfortunately, none of the business-related activities the authors included in this specific model (Snel, Engbersen, and Leerkes 2006, 302) were measured in the present study. Consequently,

28 See for example: Itzigsohn and Giorguli Saucedo 2002; Snel, Engbersen, and Leerkes 2006; Dekker and Siegel 2013; Bilgili 2014a; Schunck 2014.

Theoretical considerations taking the aforementioned results into account, it is hypothesised that economic transnationalism will neither have a significant positive nor negative effect on either of the observed dimensions of integration (HI3).

Snel and co-authors (2006, 302) find no relation between sociocultural transnational activities, e.g., frequent contact with family members in the CoO and trips undertaken to visit them, and social integration. The same holds true for Dekker and Siegel (2013, 11) in regard to the amount of time respondents spend with CoR nationals.29 However, for identificational integration Snel et al. (2006, 302) do indeed find a negative effect of sociocultural transnationalism. Regarding this second dimension of integration, a study by Braun and Müller (2012) should also be considered. While the authors mainly focus on factors determining European identification amongst intra-EU migrants, they also compare this identificational integration in a supra-national context to respondents’ identification with the CoR and CoO. In the above-mentioned text, Snel and co- authors use a composite indicator which combines cross-border contact with visits to the CoO and membership in a CoO organisation to measure the impact of sociocultural transnationalism on identificational integration. Braun and Müller, in contrast, include two specific indicators of transnational communication with family and friends in their analysis. In doing so, they find no significant effect of these variables on the identification with the CoR, contrary to what the results of Snel et al. (2006) would suggest for this type of integration. This means, while the study that uses a broader operationalisation suggests that the impact of communication and transnational mobility might differ between the dimensions of integration, this interpretation is not supported by the more specific analysis of Braun and Müller. Hence, it is hypothesised for all three indicators of integration that neither transnational communication (HI4) nor visits abroad (HI5) are significantly related to them.

Finally, it is to be expected that the origin of migrants and the country they settle in influence their degree of integration in the CoR. Relevant factors that have been discussed in this regard include geographic distance between countries, cultural differences, size of established migrant communities in the CoR, and differing legal residency statuses (Waldinger 2008, 24; Schans 2009, 1178; Carling and Hoelscher 2013, 955). However, as in the case of transnationalism, no well-founded hypotheses can be formulated due to the lack of a specific frame of comparison for migrants of both nationalities in the three CoRs. Moreover, there is a lack of data with regard to

29 Both mentioned studies use composite indicators of transnationalism to measure its influence on integration.

Theoretical considerations the question of whether the status as EU-citizen, or its absences, might influence the identificational integration of migrants. Nevertheless, it is likely that EU-citizens, due to their specific legal status and the corresponding rights, have a higher tendency to identify with the country they live in. Therefore, it is hypothesised that Romanian migrants identify stronger with their CoR than Turkish migrants (HI6). At the same time it is not clear if such an effect – if existent at all – would also be visible with regard to the town of residence (ToR) since the respective rights (e.g., freedom of movement, labour market access, transferability of social benefits) are regulated on a national level.

Table 1 provides an overview of all above-formulated hypotheses.

Table 1: Overview of hypotheses

Number Hypothesis

HT1.1 Men are more likely than women to engage in formal transnational activities. HT1.2 Women are more likely than men to engage in informal transnational activities. HT2 The longer migrants reside in the CoR, the less likely they are to engage in transnational practices. HT3 Respondents with higher formal education are more likely to engage in transnational activities than those with lower education. HT4 Labour market integration is positively associated with economic transnationalism. HT5 A positive assessment of the current household financial situation is positively associated with economic transnationalism. HT6 Respondents who have property in the CoR are more likely to engage in economic transnationalism and transnational mobility than those who do not have such real estate. HT7 The more friends and family members respondents have in the CoO, the higher their probability to engage in economic transnationalism.

HI1 The size of social networks in other countries is negatively related to social integration in the CoR. HI2 Identification with the CoO is negatively related to identification with the CoR. HI3 Economic transnationalism is neither significantly related to social nor to identificational integration. HI4 Transnational communication is neither significantly related to social nor to identificational integration. HI5 Transnational mobility is neither significantly related to social nor to identificational integration. HI6 Romanian migrants identify stronger with their CoR than Turkish migrants.

3 Methods, research design, and data

3.1 Methodological considerations A first and very fundamental methodological issue which should be discussed is the decision to use a quantitative approach that was taken at the beginning of the research process for this study. It has already been indicated in the preceding chapter that quantitative findings on migrant transnationalism are still limited as the majority of existing research applies qualitative methods. These studies are without a doubt of great value as they describe in detail how transnationalism unfolds on an individual level and which role it plays in the daily lives of migrants. Furthermore, qualitative findings show how transnational engagement connects individuals to their communities of origin, possibly even transforming them. However, an important limitation of many of these studies is the fact that respondents are predominantly sampled on the dependent variable (Portes, Guarnizo, and Haller 2002, 279; Schans 2009, 1167; Fauser 2017, 2-3). This means that the respective projects purposefully recruited informants who are part of transnational networks and engage in transnational practices. Consequently, their results can provide important in-depth information on transnational phenomena. However, this information is, by definition, retrieved from a small number of cases, which poses the question of its generalizability. In order to avoid misunderstandings, it should be stressed that this does not diminish the validity nor value of these qualitative findings, but is merely a consequence of the applied method which should be taken into account in the interpretation of results. In contrast, the use of quantitative data means that transnationalism cannot be described in as much detail as a qualitative case study would allow. However, the fact that the participants in the survey upon which this research is built were not sampled on any of the dependent variables (i.e., neither on aspects of transnationalism nor integration, see below for a detailed discussion) means that the sample exhibits a large degree of variation on those variables. Hence, while qualitative studies usually lack a comparison group (Faist 2012, 64), the data used here do not only include information on a large number of individuals who engage in specific transnational activities but also on respondents who refrain from them. Accordingly, to paraphrase Bilgili (2014b, 58), this data cannot only explain why migrants become transnationally active but also why they do not. Furthermore, the idea of comparison is important with regard to the cross-national character of this work. More precisely, migrants originating from two different countries are compared in three distinct CoRs, with regard to their degree of transnationalism, and how it relates to the mentioned aspects of

Methods, research design, and data integration. This analytical design requires comparable sample sizes in all six subgroups which must simultaneously show a certain variance on the variables of interest and their predictors. Furthermore, as previously mentioned, there is generally few quantitative research on migrant transnationalism, particularly of a cross-national comparative nature, in Europe. This gap is even more apparent for research relating transnationalism to integration. Finally, the results promise to be valuable not only to inform further quantitative research but also to contextualize qualitative results and provide stimuli for future research in either method or their combination. In this sense, the present work is based in the author’s conviction that both research paradigms should not be seen as competing, but as complementary to each other.

This analysis will help to narrow the discussed research gaps by building on quantitative data that were collected in three member states of the European Union using the same methodology and survey instrument. The uniformity of measurements and operationalisations in the analysis of samples from various countries constitutes a first important difference between this work and previous studies. By including migrants surveyed in Germany, Denmark, and Italy, this work will broaden the geographic scope of the research on the relation between transnationalism and integration. Specifically, there are no quantitative studies on these topics published for the latter two countries. With a focus on migrants from Romania and Turkey, the chosen design will also allow a direct comparison of migrants from two specific CoOs in the three mentioned CoRs. The inclusion of Romanians, as intra-EU migrants is another significant contribution to the field, as previous studies on the topic have included only migrants from so-called third countries. Their comparison with Turkish migrants will additionally allow for the assessment of the possible effects of European citizenship. Hence, by observing commonalities and differences between groups and national contexts, the analysis will take advantage of what Crul and Schneider called Europe’s character as “a ‘natural laboratory’ for integration processes” (Crul and Schneider 2010, 1250). As indicated above, this research design allows comparisons between migrants from different CoOs in the same CoR, and between migrants from the same CoO in different CoRs.

3.2 Analytical strategy and employed techniques The empirical part of the present work will begin with a comprehensive review of the samples’ sociodemographic composition. The respective chapter will provide a comparative overview, focussing on features shown by previous research to influence transnational practices and

Methods, research design, and data integration into the CoR. Amongst others, relevant indicators include the year and circumstances of migration, age, gender, education and economic situation of the respondents.

The analysis of transnationalism concentrates on phenomena which can be described as aspects of transnationalism from below, according to the systematisation proposed by Guarnizo and Smith (1998). This means the focus is on individual networks, personal traits, and private practices (see further details of the operationalisation in chapter 3.4.2). Furthermore, the discussion does not consider collective transnationalism. Hence, it will focus on practices which are linked to the individual, as opposed to practices which are, e.g., carried out through migrant organisations.

As discussed in the previous chapter, the present work understands transnationalism as a multidimensional phenomenon and uses a corresponding operationalisation (see below for details). While this systematisation is derived from the literature, it combines approaches which were applied to different types of datasets than the one used here. Therefore, in the first step of the analysis of transnationalism, the logic underlying the used classifications will be verified empirically. To this end, multiple correspondence analysis (MCA) will be employed in order to empirically confirm relations between indicators in the individual dimensions and subdimensions of transnationalism. Michael Greenacre has summarised the basic idea of this technique as follows: “The primary goal of correspondence analysis is to transform a table of numerical information into a graphical display, facilitating the interpretation of this information.” (Greenacre 1994, 3). In essence, correspondence analysis and multiple correspondence analysis allow for the graphical display of data included in the rows and columns of contingency tables. In line with the purpose of its application in the present case, it is underlined in the literature that MCA is an exploratory technique which first and foremost serves to describe data (Greenacre 1994, 3; Blasius 2010, 367). Its main advantage is the possibility of displaying the relation between variables in a geometric space with a low number of dimensions. In accordance with most MCA applications, a two-dimensional display is chosen in this study.30 The different values of variables, and more specifically of indicators of transnationalism, will be plotted in this two- dimensional space. Of key importance for the interpretation is the combination of the plotted points’ absolute position and their relative position to each other. First, the closer a value is

30 For examples of the practical application respectively the theoretical reasoning of the two-dimensional approach see: Braun and Recchi (2009, 90), Schneider et al. (2012, 301-02) and Blasius (2010, 370).

Methods, research design, and data positioned to the origin of the coordinate system, the weaker its relation with all other values in the plot. Second, the closer together two or more values are positioned and the further away they are from the origin, the higher their relation. To put it another way: the more indicators are apart in the coordination system, the weaker is their relation (Braun and Recchi 2009, 90-91; Blasius 2010, 367). Accordingly, a straightforward relation between two indicators of transnationalism exists if, on the one hand, their zero-values and, on the other hand, comparable higher values are located close to each other.

In the second step, the empirical confirmation of the multidimensional systematisation will be followed by an analysis of the extent of respondents’ transnational human capital, background and networks, as well as the degree to which they engage in the various practices. As mentioned, existing research mostly studies migrant transnationalism empirically as a phenomenon which unfolds within the dichotomy of CoR and CoO, and does not consider individual connections to third countries. The present work intends to address this limitation by including several indicators based on survey questions that referred to more than those two countries. To highlight the merits of this more open research perspective, two aspects, namely the geographic dispersion of respondents’ transnational networks and the range of their transnational mobility, will be discussed in detail. These specific indicators are also chosen because they allow for comparison with the findings of Mau and collaborators on geographic aspects of transnationalism amongst the general population in Germany (Mau 2007; Mau and Mewes 2007; Mau, Mewes, and Zimmermann 2008).

Finally, in the third step of the analysis of transnationalism, determinants of all five types of practices (i.e., transnational communication, economic transnationalism, transnational mobility, political transnationalism, and transnational consumption) shall be investigated. To this end, individual indicators of the different subdimensions will be merged into compound variables which will serve as dependent variables in multivariate regression analysis. These models will take not only sociodemographic variables but also the influence of transnational human capital as well as transnational background and networks into consideration.

The final empirical chapter will discuss the relation between transnationalism and integration in the CoR. As argued in chapter 2.2, migrant integration is understood as a multidimensional phenomenon as well. Regarding the categorisation of these dimensions, this work will employ

Methods, research design, and data

Esser’s proposal (1980; 2000; 2001), which differentiates between the four dimensions of cultural integration, social integration,31 structural integration, and identification. As shown in the previous chapter, most quantitative studies investigating the relation between transnationalism and integration do so by concentrating on aspects of structural and cultural integration. Far less information is available regarding social and identificational integration – which are the focus of this work – even though these dimensions are often central to public discussions about the continuous ties of immigrants to their CoO. The analysis of the individual dimensions will begin with a descriptive review of the respective integration dimension for all six samples. This will be followed by regression analyses using the indicators of integration as dependent variables, introducing aspects of transnationalism in subsequent models. It has to be stressed again, that the aim of this analysis is not to propose or test a model which could fully explain immigrant integration. Instead, the goal is to assess whether there are significant associations between integration and transnationalism and, if so, to determine whether they are of a positive or negative nature.

All statistical analyses and computations in this study were carried out with the Stata software. The maps used for the geographic depiction of respondents transnational social networks and mobility were generated with the freeware TileMill.32 The analysis used unweighted data. Details regarding the composition of the dataset are given in the next section. The following chapters will, of course, compare findings with those of earlier studies to contextualise results and highlight general tendencies. However, it should be kept in mind that such comparisons always retain a basic level of uncertainty due to differences in survey design of the used studies and the analytical strategies employed by the different scholars.

3.3 Survey design and fieldwork This work builds on data collected as part of the research project “The Europeanisation of Everyday Life: Cross-Border Practices and Transnational Identities among EU and Third Country Citizens” (EUCROSS 2016),33 which started in 2011 and concluded in 2014. The main research topics of the project revolved around the question of whether regular cross-border activities

31 Esser calls this dimensions ‘interaction’. See chapter 2.2.2 for the reasoning behind the decision to use a different term. 32 See: https://tilemill-project.github.io/tilemill/ (last visited: 2017/04/02). 33 EUCROSS has been funded by the 7th Framework Programme of the European Commission. All data, questionnaires and documentation are available through the GESIS Archive.

Methods, research design, and data influenced national and supra-national identification of EU residents. The EUCROSS survey was realised through Computer Assisted Telephone Interviews (CATI), resulting in a complex dataset that includes data on citizens of six countries, and on non-naturalised Romanian and Turkish migrants. This analysis will, however, use only data from the two migrant samples in Denmark, Germany and Italy. First, these CoRs were chosen for their different significance as migrant destinations. While Germany hosts the biggest Turkish community in the EU, Italy is a main destination of Romanian migration. In contrast, Denmark plays only a minor role as target country for migrants of either nationality (see chapter 4). Second, all three CoRs belong to the Schengen area. This means that the same opportunities and restrictions of physical mobility within this space are offered to third country nationals (i.e., the Turkish sample) residing legally within their borders. Finally, these CoRs were chosen because migrants of both nationalities were surveyed in all of them, which does not hold true for the remaining countries in the EUCROSS study.34 The inclusion of respondents from both CoOs in the surveys in Denmark, Germany, and Italy allows for the comparison of the two groups within and across all three countries.

In the migrant sample, EUCROSS targeted Turkish and Romanian adult citizens who were not born in the CoR and did not hold its nationality.35 Hence, respondents had to be at least 18 years of age and possess either of the aforementioned nationalities without being born in the surveyed country. Consequently, the samples include only migrants in the original sense of the word, i.e. people with personal migration experience. On the contrary, descendants of migrants are not the subjects of this study. The fact that only individuals who did not hold the citizenship of their CoR were included in the samples allows for a clear differentiation between migrants possessing EU- citizenship (Romanians) and those who do not (Turks).

The sampling procedure consisted of two parts. First, onomastic screening of telephone registries was performed. In this procedure, combinations of given names and surnames included in a list (in this case, the public telephone registry) are analysed and assigned to specific nationalities or ethnic groups (Humpert and Schneiderheinze 2000). For this study, this meant that telephone registries in all CoRs were scanned for combinations of given names and family names likely to belong to individuals from Romania or Turkey. The outcome of this first step of the sampling

34 Surveys were carried out in Denmark, Germany, Italy, Romania, Spain and the United Kingdom. Turkish migrants were interviewed in all countries but Spain. By definition, no data on Romanian international migrants could be collected in Romania. 35 Migrants who held the nationality of the CoO and the CoR were also not included in these samples.

Methods, research design, and data process was a list from which the field institute could draw random samples. The sampled individuals were then called and, in a second step, subjected to a screening procedure. During the initial moments of the conversation, interviewers inquired whether the person they had called belonged to the target group. The use of origin, age and nationality as sampling criteria meant that the respondents were neither sampled on a dependent nor major independent variable of this analysis (besides citizenship). In other words: the analysis is not restricted to individuals who were selected because they showed a high level of transnational activities or a specific degree of integration.

Slight deviations from the initial onomastic sampling were necessary for the Turkish sample in Italy. In this case, the total number of potential respondents identified by the procedure was comparatively low. Additionally, the field institute was confronted with a high refusal rate in this country. Although the field institute employed different strategies attempting to overcome these obstacles, only 49 respondents could be recruited in the standard CATI setting. Ultimately, it was decided to pursue a face-to-face strategy in order to complete the sample. Adequate sampling points were identified by the interviewers in cooperation with the Italian project team. The resulting interviews were mostly conducted as Computer Assisted Personal Interviews (CAPI, N = 179) and to a smaller degree as CATI via Skype (N = 22). In both modes, the field personnel recorded the respondents’ answers in the same software used for the telephone interviews. Hence, the recruitment strategy and survey mode of the Turkish sample in Italy somewhat differs from that of the other five samples which were sampled and surveyed following the standard procedure (for further details see Pötzschke 2015, 14-16). To avoid the introduction of a sampling bias in the analyses, all regression models in this work will, therefore, control for the mode of data collection (CATI, CAPI, CATI-Skype). The project aimed to survey 250 members of both target nationalities in all three CoRs. Applying the aforementioned measures in Italy, this sample size was reached in all cases. In Germany it was even slightly surpassed in the case of Turkish migrants, where 252 individuals were interviewed. Hence, the total sample of this study consists of 1,502 respondents, of whom 752 are Turkish and 750 are Romanian migrants.

The data were collected through standardized questionnaires in the Turkish and Romanian language which were administered by native speakers. The average duration of the interviews was slightly less than half an hour. For practical and economic reasons the survey could only be conducted in one language for each nationality, as which Turkish and Romanian were chosen.

Methods, research design, and data

Therefore, especially in the case of migrants holding Turkish nationality, members of ethnic and linguistic minorities who have settled in significant numbers throughout Europe (e.g., Kurds) might be underrepresented in the samples. The survey did not collect any data on respondents’ ethnicity. In accordance with the sampling criteria, the description of individuals as Romanian or Turkish throughout the text refers to the citizenship-status of the respondents and not to ethnic categories.

The following subchapter will provide detailed information on the measurement of several key variables, the operationalisation of transnationalism, and the dimensions of integration which this work will focus upon.

3.4 Measurement and operationalisation of key aspects

3.4.1 General sociodemographic and migration-related information The EUCROSS study used a complex and comprehensive questionnaire which employed both standard and innovative instruments. The measurement of several key demographic variables is straightforward and does not need much discussion. These are, for example, gender, country of birth and current citizenship. The last two variables were also vital in defining the target populations and, consequently, of particular importance for this study. Key sociodemographic aspects that are specific to migrant surveys, and were part of the questionnaire, include the year of (last) permanent settlement in the CoR and reasons for migration. In the analysis, the latter are grouped into the following four categories: work, education, quality of life, and family and love. Each of these categories was built from several questionnaire items in which the respondents’ answers were coded during the interview process. For example, answer categories under the quality of life headline included migration realised in order to enjoy a pleasant climate, to live in a culturally stimulating environment, or to gain new experiences in general. In order to reflect the complexity of social life, respondents could give more than one reason for their mobility.

The analysis considers a total of three time-related variables which are age, age at migration, and duration of stay in the CoR. For the calculation of these indicators, two items which asked for the respondents’ year of birth and year of settlement were vital. Age was computed by subtracting the year of birth from the year in which the survey was conducted. The variable indicating the duration of stay was built in a similar way, using the year of settlement instead of birth. Finally, the age at migration was determined by subtracting the year of birth from the year of settlement.

Methods, research design, and data

While all three of these variables will be discussed in the descriptive analysis, only age at migration and duration of stay will be used in the regressions (see chapter 2.5). However, when reviewing the determinants of transnational practices, it is conceivable that the effect of age (which corresponds to the sum of age at migration and duration of stay) could become dominant and superimpose effects of the two other indicators. Therefore, when effects of age at migration and duration of stay appear, additional computations will be reported which check for a possible confounding influence of the age variable.

Education was measured in terms of the highest formal educational title achieved by the respondents. In brief, participants in the study were first asked whether they received their highest educational title in the CoO or the CoR. A follow-up question then recorded their educational level in a battery specific to either country. Respondents who received their highest educational title in a third country were asked to state its closest CoO equivalent. In order to allow for the evaluation of differences in educational achievements across all samples, this information was later coded into the categories lower secondary education or less, in-between lower and higher secondary education, higher secondary education, and tertiary education. Appendix 1 shows the allocation of the country-specific items to these four categories.

With regard to personal relations and integration, the level of knowledge in the majority and main administrative language of the CoR is an especially relevant aspect. Therefore, respondents were asked how well they spoke the language in question at the time of the interview. To aid respondents with their self-assessment they were provided with the answering options: not at all, poorly, just so-so, quite well, and almost as well as native language. For the analysis, these categories are recoded into three items which are no or poor knowledge, intermediate knowledge, and good or very good knowledge. Due to a routing error in the questionnaire, information regarding the command of the CoR language is, unfortunately, missing for a part of the sample. Specifically, this mistake had the consequence that only respondents who reported friends and family abroad (i.e., in any other country than the CoR) were asked for their knowledge in the CoR language. Admittedly, as will be shown in chapter 5, this is the case for the majority of respondents. However, it does also mean that this information is systematically missing for the 16 percent of the overall sample who did not report any social contacts in other countries than the ones they were living in. Hence, these individuals are not part of transnational social networks. This, in turn, could lead to biased results of the analysis. To avoid this situation, all regressions

Methods, research design, and data which use the indicator for knowledge of the CoR language will include a variable controlling for missing answers due to the routing error.

Besides language proficiency, a second indicator of cultural integration is the respondents’ familiarity with their CoR. To assess this aspect, participants were asked: “Apart from the region where you live, are there one or more regions in [COR] that you are very familiar with – that is that you know well enough to feel comfortable in?”. The three answer categories to this question allowed participants to state that they knew no other region, one other region, or two or more other regions in the CoR.

The first indicator of structural integration which will be used in the analysis indicates whether the migrants acquired their highest educational title in their CoR. In addition, the integration of migrants in the labour market is of high interest. The corresponding variable includes five categories. The first is working, which includes both full-time and part-time occupations. Non- working includes respondents who were still in education, and individuals who stated that they were doing housework or looking after children and, consequently, did not have a paid job at the time of the interview. The categories retired and unemployed include respondents who did not participate in the labour market, because they were retirees or because they were out of work but not in education etc. at the time of the survey. The fifth category other refers to respondents who did not participate in the labour market due to other, specifically mentioned reasons, such as permanent sickness or disabilities.

The respondents’ income was not measured directly but ascertained in an indirect manner using a question which asked for an assessment of the financial situation of their current household on a five-point scale. The following answer categories were offered: we are living very comfortably on the money we have, we are living comfortably on the money we have, we make ends meet, we find it difficult, and we find it very difficult. The resulting variable is reverse coded for the present analysis, taking values from one to five. Hence, this variable describes the general financial situation of the interviewed migrants, rather than their individual income. This takes into consideration that the revenues of other household members, most notably their partners, could potentially influence the material situation in which respondents find themselves. The data also include a similar variable which asked for a retrospective evaluation of the financial situation of the household that respondents lived in at age 14.

Methods, research design, and data

An additional feature that serves as an indirect indicator of household wealth is real estate ownership. As part of the survey, respondents were asked, whether they or their partner, if they were living in a relationship, owned property in the CoR, the CoO, or a third country. While property in the CoR functions as a final indicator in the measurement of structural integration, the other two items are part of the operationalisation of economic transnationalism (see next subchapter).

As in most surveys which collect information on this issue via a single indicator, discrimination experience was recorded in EUCROSS via a subjective measurement. More precisely, respondents were asked whether they ever felt discriminated against due to the fact that they were born in another country. By default, such a measurement can only be an approximation of reality as it builds on the individuals’ perception of situations they are confronted with. As Schunck points out, instances recorded by such an item will reflect situations in which a person was objectively discriminated against, just as much as situations in which individuals believed that they were treated unfairly due to their origin, while, in reality, another aspect might have caused a specific situation or outcome. For example, individuals may not be hired for a particular position due to a lack of experience or education, but they might believe that it was due to their ethnicity (Schunck 2014, 270).

3.4.2 Aspects of transnationalism Building on the theoretical arguments presented in the second chapter, the following section will provide details on the operationalisation of transnationalism used in this work and justify the inclusion of specific indicators. As mentioned, the study will differentiate between the three principal dimensions of transnational human capital, transnational background and networks, and transnational practices. Furthermore, within the latter category, a distinction will be made between five subdimensions. The decision to use a multidimensional operationalisation is strongly influenced by earlier research which indicated that determining factors vary in strength and directionality depending on the aspect of transnationalism in question (see, e.g., Schans 2009, 1179; Schunck 2014, 251). The relation between the indicators within a given dimension or subdimension will be discussed in more detail in chapter 5. Table 2 gives an overview of the indicators used to operationalise the different aspects of transnationalism.

Methods, research design, and data

Table 2: Operationalisation of transnationalism

Dimensions Subdimensions Indicators

Transnational At least moderate knowledge of an additional language

human capital Familiarity with other countries Additional migration experience Partner lives in another country Transnational Family/friends from CoO in CoO background and Family/friends from CoO in 3rd country networks Family/friends from 3rd country living abroad Via phone and VoIP Transnational communication Via mail and email Via social networking sites (SNS) Sends money abroad Received money from abroad Economic transnationalism Property in CoO Transnational Property in 3rd country practices Trips to CoO Transnational mobility Trips to 3rd countries Political transnationalism Absentee voting Received messages from abroad Transnational consumption Watches foreign-language TV content Shopping abroad

3.4.2.1 Transnational human capital Transnational human capital is operationalised using three different aspects. The first one, knowledge of additional languages, is certainly an extremely important skill, enabling transnational activities (Gerhards, Hans, and Carlson 2014, 11; Kuhn 2011, 814-15). Depending on the level of proficiency, it allows direct communication across linguistic and cultural boundaries. In the case of migrants, it is less the question whether respondents know a second but a third language which is of interest. Accordingly, it is such knowledge which this indicator focusses on. To assess knowledge of additional languages, respondents were asked whether they had ever learned any language other than the main language of their CoO (i.e., Romanian or Turkish) or their native tongue (if it differed from that the former), and the language of their CoR. If they answered affirmatively their self-assessed level of knowledge in this third language was inquired. To this end respondents were asked how well they spoke the respective language.36 The offered answer categories were: almost as well as native language, quite well, just so-so, poorly and not at all. The variable was used to build a dichotomous indicator for the analysis. The first

36 In the following, this language is referred to as the third language. For practical reasons, this includes approx. 5 percent of the overall sample who do not consider Turkish or Romanian to be their mother tongue, and mentioned knowledge in an additional (i.e., fourth) language. If respondents mentioned more than one additional language, the follow-up question referred to the language they felt most proficient in.

Methods, research design, and data three of the original categories received the value one in this variable. The baseline (i.e., zero) includes the remaining two categories and also those respondents who indicated that they had never learned any additional languages. This means that the indicator differentiates between respondents who judged their knowledge in at least one additional language to be somewhere between an intermediate and quasi mother tongue level, and those who had no or very little knowledge of any third language.

While certainly important, language knowledge is not the only skill which helps individuals to navigate successfully in a transnational space. In fact, Koehn and Rosenau (2002, 110) identify a whole battery of aspects which can be seen as elements of transnational human capital, even though these authors do not actually use this term. Of their 22 “Dimensions of Transnational Competence”, at least one third is deeply related to cultural openness and a general ability to see, recognise, and navigate cultural differences and particularities. Koehn and Rosenau stress that individuals could possess different sets of such competencies and master them to varying degrees. In fact, the authors speak of “a continuum from incapable to proficient” (Koehn and Rosenau 2002, 116, italics in the original) in this regard. It follows that a similar continuum might be possible regarding the universality of these competencies. They might have the character of a general cosmopolitan mindset or be very specifically focused on the understanding of individual cultures and countries. Such understanding manifests itself in a broad sense of familiarity with a third country. That said, this may be considered a rather soft form of transnational human capital, especially compared to language knowledge. Such a deep understanding of another country has been captured in the survey used here by asking the respondents whether they were very familiar with one or more other countries apart from their CoO and CoR. Being ‘very familiar’ with a country was, furthermore, defined as knowing it well enough to feel comfortable in it.

Finally, previous migration experience will be considered as a third indicator in the dimension of transnational human capital. While Kuhn underlines the importance of long-term stays in other countries, she assigns the respective indicators to the category of transnational practices (Kuhn 2011, 820; Kuhn 2015, 34). However, this classification is not ideal. All other aspects of the dimension of transnational practices she mentions, such as visits to other EU countries and cross- border shopping, are referring to a specific (short) time window before the survey, and are not unlikely to be repeated in the future. The same cannot be said for long-term stays in other

Methods, research design, and data countries that the respondents have realised at some point during their life. As such prolonged stays might be associated with considerable expenses, it can be assumed that they are undertaken much less often than the mentioned practices of short-term mobility and consumption. Furthermore, as will be shown below, the present data suggest that most respondents already lived a considerable amount of time in their respective CoRs before the survey was completed (see Table 3). Both aspects suggest that long-term stays in other countries are more likely to be occasional experiences than regular practices. At the same time, such stays indicate that the respective individuals had an opportunity to familiarise themselves with other countries and cultures. Additionally, these stays might have presented opportunities to acquire or strengthen different forms of human capital, such as the knowledge of foreign languages (Gerhards and Hans 2013, 101). Hence, in this analysis, past migration experience, in the sense of having lived for a prolonged period in another country, will be differentiated from the transnational practice of international mobility. The latter include more frequent short-term visits to the CoO or third countries not entailing a change of residency (e.g., shopping trips and vacations) and are likely to occur more frequently. Migration experience, on the contrary, is considered a proxy of the accumulation of transnational human capital in the form of relevant social and cultural knowledge in the respondents’ past.

3.4.2.2 Transnational background and networks As mentioned in the theoretical discussion, the dimension of transnational background proposed by Kuhn (2011) will be amended to include border-crossing social connections of respondents. Such connections are not considered part of transnational human capital as they do neither constitute specific skills nor knowledge. At the same time, these connections in themselves are not transnational activities but facilitate the latter and are potentially strengthened by them.37 While the aspects of transnational background and networks might be more stable than transnational human capital or activities, they have a certain potential to change. Just as the citizenship status of individuals, which is part of Kuhn’s original definition, can change over

37 In a more recent application of her differentiation Kuhn also includes Eurobarometer items indicating whether respondents have friends from European or non-European countries in her operationalisation of transnational practices (Kuhn 2015, 34). It is not clear, whether this refers to contacts in the CoR of the respondents, in another country or both. In the present analysis, such contacts would only be considered transnational if the respective individuals would live in another country than the respondent. Furthermore, they would, in this case, be understood as part of the transnational background dimension.

Methods, research design, and data time, so can social networks. People might acquire new contacts or fall out of touch with old ones.

In order to comprehend the extent of respondents’ transnational social networks, the survey inquired whether they had a lot, a few, or no family members and friends in different geographic areas. The collection of this data followed, in two respects, a respondent centred approach. First, the questions referred to individuals from the country of origin or a third country, instead of asking directly about family members and friends of Turkish or Romanian nationality. This means that the given answers reflect the construction of national, cultural, and geographic groups by the respondents themselves. Thus, for a contact to be viewed as being ‘from Turkey’ it is not important whether the respective person holds Turkish nationality, but rather how they are perceived by the respondent. Second, respondents were not asked to give information on the absolute number of their contacts, but to quantify the relative size of each group.

After answering the questions regarding the existence of contacts abroad, respondents were asked to state which countries these individuals were living in. This information will be used to describe the geographic dispersion of individual networks and replaces in this work the typical focus of the transnational lens on relations solely connecting CoO and CoR.

For the assessment of respondents’ transnational social networks, it is, furthermore, considered whether they indicated that they had a partner in another country.

3.4.2.3 Transnational practices Transnational practices, which constitute the dimension of transnationalism that migration research usually concentrates upon, will further be divided into the five subdimensions shown in Table 2. While Kuhn does not distinguish categories of transnational practices in her 2011 text, she does do so in a more recent publication. In her analysis of the relation between transnationalism and European identity, she differentiates between long-term stays and short-term visits abroad, sociable interactions (e.g., socialising with foreigners), instrumental interactions (e.g., shopping abroad), and virtual communication (Kuhn 2015, 34). This classification is certainly particularly useful in the context of her research’s strong focus on non-migrant populations. Furthermore, it has to be taken into consideration that the author’s operationalisation builds on Eurobarometer data. This means that this data had been collected using survey instruments not adjusted to her specific research question and analytical strategy. While this is

Methods, research design, and data neither unusual nor a general disadvantage, it can be assumed that the categories she uses are partially determined by the nature of the dataset. For the present analysis, the set of subdimensions proposed by Kuhn is not ideal. In general, this work aims at a high degree of comparability with earlier findings in the area of migration research. Therefore, it seems more suitable to take categorisations of transnational practices proposed in this field as orientation. An additional argument supporting this assessment is the fact that Kuhn’s subdimensions do not cover a number of transnational activities which are typically examined in migration research, most notably aspects of economic and political transnationalism. Therefore, using the work of authors such as Portes et al. (1999), Itzigsohn et al. (1999), and Snel et al. (2006) as inspiration, five subdimensions of transnational practices will be examined in the present analysis. Namely, these are communication, economic activities, mobility, political participation and consumption (Table 2).

Transnational communication In order to evaluate the degree to which respondents’ engage in transnational communication, they were asked how frequently they talk to family or friends in other countries by phone or using a computer (VoIP), how often they communicate via mail or email, and how often via social networking sites (SNS), such as Facebook or Google+.38 Possible answer categories were: every day, at least once a week, at least once a month, less often, and never. Consequently, the used data allow not only the evaluation of the extent of respondents’ transnational communication but also their use of different means. By concentrating on communication with friends and family, this subdimension focusses on the social character of this practice. For the questionnaire structure this focus meant that the questions concerning transnational communication followed after the items inquiring about the existence of transnational networks and were filter-dependent upon them. This means that respondents who indicated that they did not have any friends or family members outside the CoR, did not receive these questions concerning transnational communication.

Economic transnationalism Money transfers are certainly a prominent topic within the field of transnational migration research. The act of sending money back to family and friends in the CoO (i.e., remitting) is a

38 It should be stressed that the question specifically referred to social networking sites and not to instant messaging services, such as WhatsApp or Facebook Messenger.

Methods, research design, and data particularly important transnational activity as it stabilises the migrants’ engagement with their CoO over time and distance. Numerous studies have shown that remittances are often an important additional source of income for the receiving non-migrant family members (e.g., Itzigsohn 1995, 653; Landolt 2001, 234; De Haas 2005, 1274). Two important points have to be clarified with regard to the corresponding measure in this study. First, this analysis concentrates on individual remittances and does not pertain to collective contributions via clubs, religious groups, or hometown associations. Second, the used data only contain information about the act of sending money and not the amount sent. This is due to a conscious decision in the survey design because the interpretation and comparison of specific sums across social strata and national groups would be problematic. Due to family structure, income, and education, sending 100 Euro a year might be, for example, just as challenging for some respondents as sending the same amount per month might be for others. Disentangling these and other influencing factors would have required a much more extensive measurement of this potentially sensitive topic.

Besides sending money, it is also conceivable that money transfers are realised in the opposite direction, putting migrants at the receiving end of such transactions. This is accounted for by an additional indicator which measures whether respondents received any money from someone in another country than the CoR during the 12 months preceding the survey.

Portes et al. (1999, 219) state that the onetime purchase of property in another country should not be seen as a transnational activity since it is – by definition – not performed on a regular basis. Nevertheless, possession of real estate in the CoO or a third country will be considered an additional aspect of economic transnationalism besides remitting and receiving money. This is due to the fact that possession of real estate in another country can be seen as an indirect indicator of persistent transnational engagement. First, property in another country means by definition that its owner has a lasting material bond to the country in which it is situated. It can be assumed that most persons in this group will consequently also have a certain interest in the economic, political, and social developments in said country. Second, real estate is tied to a number of formalities and obligations, most notably of a fiscal nature, which suggests that the owner of such real estate has to engage in cross-border dealings on a regular basis, even if that is only once a year. Finally, the existence of real estate abroad will most likely have the consequence of at least occasional interactions between migrants and individuals in those countries who are in charge of the owners’ property in their absence. If respondents actually are engaged in housing

Methods, research design, and data development, this contact may be very frequent at times. Finally, the existence of property in another country can also generate additional income, when it is rented to third parties (L. Smith and Mazzucato 2009, 665). The significant role that dwellings and particularly their building processes can play within a transnational social space have been discussed by several authors (Pauli 2008; Dalakoglou 2010; Erdal 2012). In this context, the potential importance of real estate in the transnational space is summarised by Erdal as follows:

“[E]ven when the migrants are not staying in their house, its relational signiﬁcance remains. Someone may be looking after the house, either living in it or passing by regularly. The house is a point of reference in conversations across transnational space.” (Erdal 2012, 637)

In light of these considerations, respondents were asked whether they or their partner, if applicable, owned real estate not only in the CoR but also in the CoO or a third country.

Transnational mobility As discussed, social connections across borders are an important aspect of transnationalism. These connections manifest themselves in networks which allow for regular cross-border communication. The fact that the individuals who form part of these networks are located in a number of different geographic locations and at distances which do not allow them to just casually visit each other is a defining characteristic of such cross-border ties. While long-distance communication plays an important role in the formation and maintenance of long-distance social networks, Urry (2007) underlines that they also require physical co-presence of the involved individuals from time to time. The author mentions different processes which can trigger movement to allow for such meetings. Most important in the migration context are formal events related to social and familial life (such as weddings, birthdays, and funerals), and social obligations of a less formal manner which still revolve around the expectation of being present and willing to spend ‘quality time’ with each other. Urry points out that such meetings are “necessary for sustaining trust and commitment that will have to persist during periods of distance and solitude” (Urry 2007, 233). As O'Flaherty et al. (2007, 820) put it, mobility also plays a role in transnational social networks which is paralleled by long-distance communication. While the former allows for the occasional physical co-presence as an important strengthening factor for network connections, the latter allows the continuance of these connections during the everyday reality of physical and geographical distance. However, in addition to movements related to social meetings, there are a number of other factors which can induce mobility. In the

Methods, research design, and data context of migration, these could, for example, be related to legal requirements, for instance in cases in which passports can only be renewed in the CoO, or educational necessities, e.g., if titles cannot be authenticated by embassies but only by the issuing institution. Another example would be visits to the CoO in order to participate in an election for which the use of absentee ballots is not possible. Lastly, on a more generally level, cross-border mobility is often also connected to leisure time activities (such as shopping and vacations), and can be caused by professional necessities. These types of mobility have recently been studied with regard to the transnationalism of a largely sedentary population by Mau and his co-authors (Mau 2007; Mau and Mewes 2007; Mau, Mewes, and Zimmermann 2008).

In contrast, migration research usually employs a more restrictive geographic focus by concentrating on mobility which occurs between the CoR and CoO. The present work will contribute to the narrowing of this research gap by taking visits to third countries into account as well. To measure the latter, respondents were asked how often they had visited another country than CoO and CoR during the 24 months preceding the survey. While the duration of such stays was not inquired, respondents were instructed only to count those trips which included at least one overnight stay. A follow-up question asked respondents to list all the visited countries. Similarly, the survey item that inquired about the number of visits to the CoO also referred to travels during the last 24 months.

Political transnationalism Political transnationalism is without a doubt a very complex and controversially discussed topic. A range of different activities has been studied as indicators of this phenomenon. For example, Østergaard-Nielsen (2003, 762) includes in her definition the active support for political actors in the CoO, participation in political debates of the CoO, and voting in CoO elections. Other authors include, in some cases in addition to the former, news media consumption, party membership, monetary donations to political parties, participation in CoO oriented charities, and financial support for civic projects in the community of origin (e.g., Guarnizo, Portes, and Haller 2003, 1223; Snel, Engbersen, and Leerkes 2006, 292). Pertinent aspects explored in this context also include multiple citizenship and the efforts of sending countries to exert influence over emigrants (e.g., Jones-Correa 2001; Bauböck 2003; Serrano Carrasco 2003; Pötzschke 2016). Hence, participation in elections from abroad is only one of a number of activities which could be used to measure transnational political engagement. However, in this study, no additional indicators are

Methods, research design, and data available due to the fact that the EUCROSS research project surveyed both migrants and non- migrants, and the questionnaires for all samples were kept as uniform as possible. Therefore, in using a single indicator of political transnationalism this study’s measurement is much less complex. This also means that the analysis of this aspect is somewhat limited and likely to underestimate the extent to which respondents engage in transnational political activities.

Transnational voting behaviour was measured by the question “Did you vote in the last general elections in [Turkey/Romania] in [month and year of the election]?”. For Turkey, this question referred to the general elections in June 2011, and for Romania, to those in November 2008. With respect to the transnational nature of the participation in CoO elections, it is important to mention that a little over three percent of all Turkish, and exactly 18 percent of Romanian respondents, settled in their CoR after the relevant election date or were younger than 18 years at this time and therefore could not vote.39 An indicator controlling for this fact will be included in the regression analyses.

Transnational consumption The last subdimension of transnational practices investigated here centres on mundane practices and more specifically on migrants’ cross-border consumption of goods and media without being physically mobile. Quantitative research which takes aspects of transnational consumption into consideration mostly focusses on the degree to which migrants use CoO media. Snel et al. (2006), for example, include in their analysis the question whether respondents read CoO newspapers, while Bilgili (2014a) presents findings on the consumption of art and media which originated in the same country as the migrants themselves. Studies applying qualitative methods also investigate media consumption, mostly concentrating on the degree that migrants watch CoO satellite TV (e.g., Aksoy and Robins 2002). Other researched aspects include culinary habits and home decoration (Üçok 2006), or transnationally linked local ethnic economies in the CoR consisting, for instance, of shops, restaurants, and tea houses (Ehrkamp 2005). Thus, what can be observed in the literature is, on the one hand, a strong focus on media consumption and, on the other hand, on practices which are directly related to the respondents’ CoOs or their origin culture in a broader sense.

39 The minimum voting age in both countries is 18 (CIA 2017).

Methods, research design, and data

As it is one of the objectives of this study to widen the analytical focus with regard to transnational practices and their geographic scope, the indicators used to measure this type of transnationalism differ from those described above. More precisely, the analysis will focus on cross-border shopping, the consumption of foreign-language TV content and the extent to which respondents are exposed to electronic messages originating from other countries.

Cross-border shopping, in the sense of purchasing goods from afar without engaging in physical mobility, is a form of marked bound cross-border activity that has, during the last two decades, become much more accessible to larger parts of the population in many countries worldwide. This is mainly due to the spread of the internet. Regarding migrants, it is of interest whether and to what degree they make use of the increased possibility of consuming products from their CoOs without relying on an ethnic economy in their residential area. The specific item used to gather this information read: “Thinking about the last 12 months, have you purchased any goods or services from sellers or providers who were located abroad? That is, for example, via websites, mail, phone, etc.” This was followed by a second question which inquired about the country that providers of the purchased goods or services were located in.

In general, watching foreign-language TV or consuming such content via the internet, can be considered a virtual way of transcending geographical distances as well as political and cultural frontiers. As Urry points out, television has resulted in a “’shrinking’ of the distances of time and space” (2007, 169) since the second half of the 20th century (also see: Aksoy and Robins 2002, 9). In the context of this work, this reduction of distance is of interest as content traditionally diffused via television allows consumers to embark on what Urry calls a form of ‘imaginative travel’ (2007, 169). The measurement of the consumption of TV content used here differs in two important ways from the one in the above-mentioned studies. First, the measurement in this survey did not focus on content produced in the CoO but rather asked how often the respondents consumed TV content in a language other than that of their CoR and CoO. While the question included content that used subtitles, it excluded dubbed content. The focus on language rather than on the (perceived) country of production means that a broader approach is taken than in the literature. The exclusion of CoO language content means, furthermore, that this measurement implies a virtual crossing of cultural boundaries rather than a possible urge to (re)connect with the CoO from afar. Second, the used question did not specify the type of device that TV content was consumed on. Therefore, the survey took the growing tendency to watch and stream online

Methods, research design, and data content, as well as the increased availability of foreign-language TV shows and movies into consideration.

The last indicator of this dimension reports on the quantity of electronic messages respondents received from outside their CoR during the 12 months preceding the survey. The respondents were asked to state an approximate percentage in this regard. In contrast to the communication indicators reported earlier, this measure did not presume that the respondents engaged in a conversation or reacted to these messages. Instead, it also refers to the somewhat passive consumption of newsletters, for example. Furthermore, again in contrast to the communication variables, it was not limited to messages received from family members and friends, but also included messages which could have been sent by less familiar acquaintances or entities such as news agencies. It should, however, not go unmentioned that messages of friends and family abroad have most likely also been considered in the answers many respondents provided to this question.

3.4.3 Social integration In this study social integration is operationalised as the total number of friends and family originating from the CoR, that each respondent has in the CoR (e.g., friends and family of Danish origin in Denmark).40 Hence, this variable is equivalent to those that measure the size of transnational networks. This means that this variable likewise used the answer categories none, a few and a lot.

Respondents who are in a romantic relationship with a person from the CoR are also identifiable in the EUCROSS data. However, this information is not incorporated into the indicator of social integration but will be treated as an independent variable. While some authors argue that intermarriage could be seen as a sign of cultural and social integration (e.g., Alba and Nee 2003, 90; Braun and Müller 2012, 265), others demand a more cautious assessment. In this sense, Song argues that it should not be assumed “that an interracial partnership is automatically devoid of prejudice or racism within the couple relationship or the wider family network (or indeed the wider society).” (Song 2009, 341). While the author refers to ‘interracial partnerships’ in this citation, the observation she makes applies to the situation of migrants in general. In other words, it should not be assumed that migrants will automatically be integrated into social groups

40 The wording of the question was: “Please think about all family members, in-laws and friends you have who live in [CoR]. I would like to know, how many are from [CoR].” The answer categories were read to the respondents.

Methods, research design, and data consisting largely of CoR individuals which their partner is part of. Such an assumption is also problematic as it, more generally, presupposes that the CoR partners themselves are integrated into such CoR networks. Furthermore, Rodríguez-García et al. argue, based on their qualitative data, that in some cases the CoR individual “integrates into the sociocultural world of the immigrant spouse” (Rodríguez-García et al. 2015, 233). It seems logical to assume that this would not only apply to married couples (i.e., spouses) but also to non-married partners. On a more general level, the quantitative analysis of these authors indicates that the relation between intermarriages and integration is likely to differ depending on the dimension of the latter. While they, for example, report overall positive associations between intermarriage and the number of CoR nationals in migrants social networks, they find no relation between this aspect and CoR language acquisition, if standard socio-economic controls are applied (Rodríguez-García et al. 2015, 237). However, the quantitative results of this study should be interpreted with care due to the small number of observations.41 Taking similar findings into consideration, Alba and Foner (2015a) state – somewhat more cautiously than the above-cited text which was also co-authored by Alba – that unions between migrants and CoR individuals “will be generally associated with integration into the societal mainstream” (Alba and Foner 2015a, 51, italics added). This assessment suggests that the indicators of social integration and the existence of a CoR partner should be treated separately in the analysis. On the one hand, a possible positive relation between both indicators could mean that migrants come into more contact with CoR nationals through their partner and his or her network. On the other hand, it could also mean that migrants who have a higher number of CoR friends in the first place find their partners in or through these networks. Hence, the direction of a possible effect cannot be determined on a general level.

Furthermore, it could be assumed that social contacts to people of CoR origin are also influenced by social relations with individuals from the CoO or migrants from other countries. In fact, a negative relation in this regard results from of the above-mentioned original argument of Esser, whereby social integration could, besides empirically insignificant exceptions, only be successful in the form of assimilation. Accordingly, he expects that the integration in a co-ethnic group would ultimately hinder the integration in the mainstream society of the CoR (Esser 2001, 23, 41). To control for such negative or possibly positive effects, the analyses will also include the

41 The sample consisted of 94 migrants residing in Spain, of whom 58 were in relationships with Spanish-born individuals while 36 had a partner who was born in the same country as them (Rodríguez-García et al. 2015, 227- 28).

Methods, research design, and data number of friends and family members of CoO and third country origin in the CoR as independent variables.

3.4.4 Identificational integration As discussed, identificational integration is the second dimension of integration focussed on in this work. Schneider et al. argue that “identities […] defined as ‘labels’ for belonging to certain categories” (2012, 286) involve three different perspectives that must at least partially overlap in order for the use of such labels to be perceived as legitimate. These are “the self-ascription of the individual, the habitus of the category (or the group itself) and the ‘outside world’ (non-group members).” (2012, 286). Due to the nature of the data used here, this present work will limit itself to the description and discussion of the first of these aspects, i.e., the degree to which individuals see themselves as part of a specific group. This means that neither specific behaviour nor the confirmation of identifications by other individuals will be part of the analysis. Consequently, it has to be stressed that it is not the formation or variation of identities in their entirety which is in the focus of this research. Instead, the discussion will concentrate on individual identifications and the degree that they are influenced by transnationalism.

Furthermore, research has shown that individuals usually identify with more than one group and that the degree of these identifications can differ between the local and the national level (Díez Medrano and Gutiérrez 2001; Schneider et al. 2012). Therefore, this work will not only investigate the identificational integration in the country of residence but also in the town of residence (ToR). The necessary data were collected by asking how much respondents agreed with the statements “I feel Danish/German/Italian” and “I feel as a citizen of the town where I live”. The level of agreement was measured using a five-point rating scale, where one meant strongly disagree and five meant strongly agree. To investigate the possible influence of the identification with the CoO, the analysis will also include a similar variable based on the assessment of the statement “I feel Turkish/Romanian”. The identification with the CoO is, however, not considered an aspect of transnationalism in this study, but rather a feeling of ethnic or national belonging (see Bilgili 2014b, 174).

4 Sociodemographic description

4.1 Turkish and Romanian migrant population in Denmark, Germany, and Italy With its focus on Turkish and Romanian migrants, this work takes into view individuals from two major origin countries of immigrants within the European Union. According to the EU’s statistical office, 7.5 percent of all foreign citizens living in EU member states in December 2008 held Turkish nationality, followed by 6.2 percent who were Romanian citizens. Thereby, people of these nationalities constitute the two biggest groups of foreign residents within the EU (EUROSTAT 2011, 74-75). In combination with these two CoOs, the discussion will concentrate on three particular CoRs in which the relative size of these two groups stand in different relation to each other and the total population.

As mentioned in the theory section of this text, previous research has shown the effect of contextual factors on transnational activities and the degree of migrants’ integration in the CoR. Factors which are assumed to influence one or both of these aspects are the geographic distance between the CoR and the CoO, structural conditions the migrants encounter in the CoR, and the size of the co-ethnic community in the country (Morawska 2002, 163). Relevant in this regard is not only the absolute number of migrants from a specific CoO in a CoR but also their communities’ relative size in comparison to other migrant groups and to the general population. Both of these aspects might affect their ability to organise, and their active and passive presences in public debates. Tubergen, Maas, and Flap point out that two possible effects of co-ethnic community’s size are discussed in the literature. First, a higher number of co-ethnics in the CoR could be an advantage for immigrants, as it enhances the probability of profiting from an existing network which is ready to assist newcomers. From this point of view, co-ethnic groups are seen as a source of ‘ethnic capital’ (Tubergen, Maas, and Flap 2004, 710). However, whilst acknowledging this potentially positive effect, Esser argues that the very same networks and the opportunities they provide could lead immigrants into a social mobility trap, meaning that they would only integrate into the established ethnic community in the CoR, while social integration in the CoR in a broader sense would be limited. This process would ultimately lead to segmentation (Esser 2001, 41; Faist 2013, 1642). The second argument summarised by Tubergen and co-authors, stresses that the bigger an ethnic migrant community in a CoR is, the more visible it is to the general public. Consequently, it would then be at a higher risk of being

Sociodemographic description perceived or portrayed as a cultural or economic threat by members of the dominant ethnic group in the CoR (Tubergen, Maas, and Flap 2004, 710). In light of these conflicting arguments, this chapter begins with a short overview regarding the size of the Romanian and Turkish migrant communities in their three CoRs at the time of this survey.

With nearly one million individuals, Romanian citizens constituted the single biggest immigrant group in Italy at the end of 2012. In fact, more than 21 percent of all foreign nationals living in Italy held Romanian citizenship, which corresponded to 1.6 percent of the overall population. In contrast, the total number of Turkish residents was comparatively low, as this group was comprised of less than eighteen thousand individuals, amounting to only 0.03 percent of the Italian population (see Appendix 2).

In Germany, these proportions were reversed. With almost 1.6 million individuals, Turks constituted the biggest group of foreign citizens, amounting to more than 1.9 percent of the overall population, and approximately 22 percent of all foreign nationals. Although Romanians were the sixth biggest group of foreigners in Germany in 2012, their number was much smaller in comparison and amounted to 0.3 percent of the country’s overall population (see Appendix 2).

Denmark offers a somewhat different picture with regard to both nationalities. Although Turkish citizens also constituted the biggest group of foreigners in the country, their numerical dominance was far less pronounced. With approximately twenty-nine thousand individuals, they constituted 7.7 percent of the foreign and 0.5 percent of Denmark’s total population. At the same time, there were also more than twelve thousand Romanian citizens living in the country, which means that they constituted 3.3 percent of the foreign population (see Appendix 2). Therefore, while the differences in the size of both groups were considerable, they were – in absolute and relative terms – much closer to one another than in the other two countries.

4.2 General migration patterns, age, gender, and migration motives The different sized communities of migrants from Romania and Turkey in the three countries under observation are a consequence of distinct migration patterns in recent history which have linked both these countries and their nationals to Western Europe.

As discussed, people of Turkish origin constitute the biggest group of foreign residents in the European Union. Their immigration started in earnest in the 1960s when Northern European

Sociodemographic description countries were in dire need of additional industrial workers. This labour related migration of the 1960s and early 1970s changed its character to mostly family oriented migration in later years (e.g., family reunification or marriage migration). Germany is by a large margin the Western European country which has attracted the most Turkish migrants since the middle of the 20th century (Crul and Vermeulen 2003; Karakaşoğlu 2008; Oğuz 2012). Large-scale migration from Romania to Western Europe, on the other hand, did not start until the late 1980s. The biggest share of these migrants went to Germany in the direct aftermath of the socialist regime’s end. The direction shifted in the mid-1990s with increasing irregular migration to Spain and Italy, which both had a comparatively high tolerance towards irregular migration at the time, including sporadic legalisation campaigns (Del Boca and Venturini 2005, 316; Engbersen et al. 2010, 9; Anghel 2013, 3). Finally, destinations diversified when Romanians were allowed visa-free travel into and within the Schengen space (2002), followed by the country’s accession to the European Union in 2007. Nevertheless, Italy still retains its status as the most important destination for Romanian immigrants (Sandu 2005, 39-40; Ban 2012, 132; Anghel 2013, 6-8).

The survey data used in this analysis pertains only to migrants who did not naturalise in the country they migrated to. It cannot be determined whether and how these respondents differ from their naturalised compatriots who are not included in this study. Consequently, it is not possible to derive general findings on the migration history of Turks and Romains to Denmark, Germany, or Italy from this data. However, findings previously presented by a wide range of researchers constitute the context in which the data should be seen. In the following, it will be briefly shown that some historical patterns identified in the literature are also visible in the used data. To this end, Figure 1 and 2 show the cumulative growth of the samples for both nationalities by CoR across time.

Sociodemographic description

Figure 1: Cumulative relative growth of samples by year of settlement in the CoR – Turkish migrants

100 90 80 70 60 50 40 30 20 10 0

1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2012 Percentage ofsamplethe (cumulative) Year of settlement TU in DK TU in DE TU in IT

EUCROSS, 2016. N=752.

While Figure 1 shows that the first Turkish respondents came to Germany42 in the early 1960s, the first significant increase in migration can be seen between 1965 and 1975. The same holds true for the Turkish sample in Denmark. This means that these increases fall exactly in a period in which political measures in both countries tried to counteract labour market shortages by allowing, encouraging, or organising the recruitment of foreign workers (Icduygu, Sirkeci, and Muradoglu 2001, 44; Pedersen and Smith 2001, 3; Nannestad 2004, 757; Oğuz 2012, 27).43 In fact by 1973, when these policies were stopped due to effects of the oil crisis on Northern European economies, 23 percent of the Turkish sample in Germany and seven percent of the sample in Denmark had already arrived at their respective CoR. The fact that both samples showed a continuous and in the case of Germany even accelerated growth pattern in the following years corresponds to broadly published findings. While Northern European countries originally intended to end migration from Turkey when they halted their “guest worker” policies, the actual result was mainly a change in its character. Deprived of the possibility of return at a later stage, many migrants who may have initially intended to return to Turkey within a few years after their first arrival now decided to stay on in Germany and Denmark. As a direct

42 If not further specified, the term Germany always refers to the Federal Republic of Germany throughout the text. Hence, the term excludes the German Democratic Republic before 1990 but includes the newly formed East-German federal states thereafter. 43 While a very formalised guest worker scheme based on bilateral contracts with sending countries was in place in Germany, the guest worker system in Denmark was less structured by the state and mainly consisted of a predefined framework in which employers could recruit workers from abroad (see Herbert (2003, 203-5) for Germany and Christoffersen et al. (2014, 239-40) for Denmark).

Sociodemographic description consequence, family reunifications became a much more prominent migration pattern since those Turks already living in Northern Europe started to invite their families to join them in considerably higher numbers than before (Herbert 2003; Karakaşoğlu 2008, 1055; Schunck 2014, 182). In accordance with these findings, bivariate analysis of the relation between different migration motives and the year of settlement in the CoR show that Turkish respondents who settled later in Denmark and Germany were significantly less likely to give work-related reasons for migration (Appendix 3). The opposite holds true for family-related migration motives, which were significantly more likely to be given by members of both samples who migrated later. This is consistent with the findings in the literature on Turkish migration, as it suggests that the growing importance of family-related migration in Northern European countries superseded work migration over time. The sample in Italy does not show any relationship between the year of settlement and these migration motives.

Turkish respondents started to arrive in Italy at a later point compared to the other two countries (Figure 1). This is hardly surprising, as Italy did not participate as a receiving country but rather as another sending country in the “guest worker” migration scheme. However, this status gradually changed from the mid-1970s onwards due to increasing wages in Italy, as well as the increasingly restrictive migration policies in Northern European receiving countries (Del Boca and Venturini 2005, 321). Accordingly, while the first migrants of the Turkish sample arrived in the early 1980s in Italy, it took until the second half of the decade for their arrivals to become more frequent.

100

Sociodemographic description

Figure 2: Cumulative relative growth of samples by year of settlement in the CoR – Romanian migrants

100 90 80 70 60 50 40 30 20 10 0 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2012

Percentage ofsamplethe (cumulative) Year of settlement

RO in DK RO in DE RO in IT

EUCROSS, 2016. N=750.

The described general migration pattern of Romanians to Western European countries is also reflected in the analysed samples, as can be seen in Figure 2. Besides a small number of early immigrants in Denmark and Germany, respondents only started arriving to the surveyed countries in larger numbers in the decade before the end of the Romanian socialist regime. In fact, the most pronounced growth in the time between 1985 and 1995 can be seen in Germany, where more than one third of the sample arrived in these ten years alone. In total, nearly half of the Romanian sample in Germany had arrived by 1995. Again this means that the sample provides a good reflection of historical movements. In fact, already during the Cold War but particularly after the end of the socialist regimes in Eastern Europe, Germany experienced significant immigration from people whose ancestors were German emigrants. These individuals were therefore considered Germans according to paragraph seven of the German citizenship law (BMJV 2015).44 As a consequence of this ethnic understanding of nationhood, specific laws and regulations significantly lowered immigration obstacles not just for ethnic Germans but also for their non-German spouses and close family members.45 While Romania was, in quantitative terms, not the most important country from which this migration movement occurred, statistics show that nearly one fifth of the 4.5 Million individuals who immigrated to Germany between 1950 and 2012 under these regulations, came from Romania (Worbs et al. 2013, 28).

44 In legal terms, these migrants are referred to as Aussiedler or Spätaussiedler. 45 The specific regulations and requirements for accompanying non-Germans were changed various times after the Second World War. A brief overview can be found in Worbs et al. (2013).

101

Sociodemographic description

Consequently, personal networks between citizens and residents of both countries were established or strengthened, which later migrants, who were not themselves ethnic Germans, could profit from. As Sandu (2005, 39) points out the migration of ethnic Germans and their families dominated emigration from Romania in the first half of the 1990s and slowed down thereafter. While the bivariate analysis does not show a negative relation between family migration and later years of settlement in Germany, it does show a positive significant relation with work migration, indicating that later arrivals in the country were more likely to state labour related migration motives (Appendix 3). The importance of the Romanian community in Germany for the early migration to other European countries is stressed by various authors, who argue that these networks allowed many Romanians to procure tourist visas in order to enter the Schengen space and to participate in the irregular labour markets of various EU countries (Potot 2010, 257-58; Anghel 2013, 6).

The main growth of the Romanian sample in Italy took place in the decade between 1995 and 2005, in which over three quarters of the respondents settled there (Figure 2). This is in good agreement with the previously discussed research that showed how Italy had become the leading destination country of Romanian migration during the late 1990s (Stan and Erne 2014, 29).

The settlement pattern of the Romanian sample in Denmark strongly reflects the institutional inclusion of their CoO in the European common market and the European Union. The data show, that nearly eighty percent of the interviewed Romanians in Denmark arrived after 2002, i.e., after their country joined the Schengen space and visa-free travel became possible for Romanian citizens. Albeit the sample continued to grow thereafter, only 38 percent of the respondents settled in Denmark before the year their country joined the European Union (2007). While more than a quarter of the respondents arrived in Denmark between 2007 and 2008, one third of the sample did not settle in the country before 2009, which was the year in which Denmark granted Romanian citizens unrestricted access to its labour market (European Commission 2011, 4). In accordance with the mobility opportunities related to the accession to the European Union, the bivariate analysis shows a significant positive relationship between a later year of settlement and work migration, as well as with educational migration of Romanian respondents to Denmark. However, quality of life and family-related reasons were significantly less likely to be cited by later arrivals (Appendix 3).

102

Sociodemographic description

For both nationalities, there is one CoR to which large parts of the samples arrived only after the turn of the millennium. These groups are Romanians in Denmark, of whom 85 percent settled in the country in 2000 or later, and Turkish migrants in Italy, of whom 74 percent did so. The latter group is also the only one which includes a noteworthy share of respondents (17 percent of the sample), which arrived within the 24 months directly preceding the survey. All respondents of the other two Turkish samples and at least 93 percent of all Romanian samples had arrived in their CoR by 2010.

Table 3 provides a summary and extension of the above-made arguments. The presentation is restricted to median values for the duration of migrants’ stay in the CoR, age at migration, and age. Median values are given because mean values are in some cases strongly influenced by outliers. When grouped by CoR, it can be seen that the differences in duration of stay are less pronounced amongst the Romanian versus the Turkish subsamples. The median value of the years Romanians spent in their CoR prior to the survey is lowest in Denmark followed by Italy and Germany. The difference between the lowest and the highest value is only 10 years. However, among Turkish migrants the difference amounts to 21 years, with Turks in Italy showing the lowest value and those in Germany showing the highest value.

Table 3: Basic demographic information

Turkish migrants in … Romanian migrants in …

Denmark Germany Italy Denmark Germany Italy Duration of stay in CoR (median in years) 19 28 7 5 15 12 Age at migration (median in years) 21 19 24 25 33 30 Age (median in years) 40 46 32 30 47 43 Gender (female in %) 47 56 44 41 57 60

EUCROSS, 2016. N = 1,474.

In general, Turkish nationals migrated in their early adulthood, with their median age at migration varying between 19 and 24 years (second row in Table 3). The fact that more than every third Turkish respondent in Germany, and more than one in four in Denmark, came to their CoR when they were younger than 18, is another indication that overall migration patterns are partially reflected in the samples (figures not included in the table).46 It is very likely that these individuals did indeed enter these countries through the respective family reunification schemes. In comparison, Romanian respondents tended to migrate somewhat later in life, i.e., in their late

46 The respective percentages in the other groups vary between two percent for Romanians in Denmark and 12 for Turks in Italy.

103

Sociodemographic description twenties and early thirties. Regarding the median age at migration, differences between both migrant groups residing in the same country are smallest in Denmark and biggest in Germany.

Amongst both nationalities, those migrants who were interviewed in Germany lived the longest in their CoR, and were, in comparison, the oldest. The difference of median age between both samples is furthermore only one year, with Turkish migrants in Germany having a median age of 46 and Romanians of 47. Besides discussed commonalities, the values for respondents of both nationalities in Germany also show an interesting difference: while Turkish individuals migrated at the lowest age, the contrary is true for Romanian respondents, when compared to their compatriots in Denmark and Italy.

The gender distribution is fairly balanced across all samples. There are slightly more women than men in the Romanian samples in Germany and Italy, and in the Turkish sample in Germany. In the other three samples, the inverse is true. This means that women are slightly overrepresented in both samples in Germany and slightly underrepresented in both samples in Denmark. In Italy, no such consistency regarding the gender distribution can be observed.

While some general findings regarding the respondents’ migration motives are reported above, Table 4 offers a more comprehensive overview. In order to reflect the complexity of social life, respondents could give more than one reason for their mobility. However, only 14 percent of the Romanian and five percent of the Turkish migrants did so.

Table 4: Migration motives (percent)

Turkish migrants in … Romanian migrants in … Denmark Germany Italy Denmark Germany Italy Work 27 21 46 51 32 64 Education 1 3 26 26 3 2 Quality of life 6 3 7 18 15 10 Family and love 70 73 26 14 54 33 EUCROSS, 2016. N = 1,484.

For respondents in all six subsamples, either work or family-related motives were most important for the decision to migrate to another country. The phenomenon of work migration is most pronounced amongst Romanian migrants in Italy, where nearly two thirds of the sample engaged in it. The same holds true for roughly half of the Romanian respondents in Denmark and of the Turkish respondents in Italy. For the respondents in the remaining three samples (Turks in

104

Sociodemographic description

Denmark, Turks and Romanians in Germany), work motives were second most important preceded only by family and love.

Only in two samples, Turks in Italy and Romanians in Denmark, educational reasons were crucial in the migration projects of a considerable number of respondents. Approximately one quarter of all respondents in both groups had corresponding migration motives. As mentioned before, these two samples also showed the lowest median age and duration of stay in the CoR.

While Turkish nationals rarely stated quality of life reasons as motivators for their migratory projects up to two out of ten Romanians did so. The fact that the lowest percentage is shown by those who went to Italy and the highest by residents in Denmark, with Germany in-between, suggests that these respondents did not primarily base their decision on climatic considerations.

4.3 Education, labour market participation, and economic situation The level of formal education an individual possesses is identified as a key aspect, both in integration and transnationalism research. In the former, the argument is that formal education has a major influence on structural integration because it is central to the labour market positioning of individuals in European societies (Esser 2001, 52). While the possession of a specific educational title alone might not guarantee equal access to a desired position on the labour market, the lack of educational capital can preclude this access altogether. Additionally, Berry (1997, 22) argues that a higher level of education can reduce the stress caused by the adaption to a new environment after migration, thus facilitating the integration process. As mentioned above, empirical studies have furthermore shown that higher levels of formal education are also positively associated with transnational involvement.

105

Table 5: Highest educational title (percent) Turkish migrants in … Romanian migrants in …

Denmark Germany Italy Denmark Germany Italy

Total* CoO CoR Total* CoO CoR Total* CoO CoR Total* CoO CoR Total* CoO CoR Total* CoO CoR Lower secondary 40 47 30 49 54 37 34 46 9 3 4 0 4 3 5 2 0 9 education (or less) In-between lower and higher 5 2 10 11 4 32 4 5 0 9 9 11 11 9 23 9 11 0 secondary education

Higher secondary 29 37 14 28 34 10 20 22 16 28 38 6 41 47 10 60 65 40 education

Tertiary 20 13 36 9 8 14 41 27 74 59 49 82 42 40 51 28 24 47 education

N 250 164 73 252 184 63 250 169 70 250 176 72 250 209 39 250 204 43 % of total N 66 29 73 25 68 28 70 29 84 16 82 17 EUCROSS, 2016.

* The totals also include the small number of respondents who specified that they received their highest educational title neither in their CoO nor in their CoR. Totals might not add up to 100 percent due to missing values.

Sociodemographic description

Table 5 shows the distribution of educational titles for all samples on a country of residence level. Three values are given in each case: the first column (‘total’, highlighted in light grey) contains the total percentage of the sample that obtained a specific level of education, while the second (‘CoO’) and third (‘CoR’) column show the percentages of respondents who received their highest educational qualification in their CoO or their current CoR, respectively. In each column, the cell with the highest percentage is highlighted in dark grey. The totals in the first column also include respondents who were educated in a third country. The differentiation of educational titles awarded in the CoO and CoR is relevant since the latter can be seen as an indicator of structural integration.

On an aggregate level, clear differences between migrants of both nationalities under investigation can be observed. An overwhelming majority of Romanians (86 percent) reported educational titles in the upper two levels. This means that only around one in ten Romanian respondents did not obtain higher secondary education. However, the same holds true for nearly half of the Turkish respondents.

The two groups with the lowest formal education on average are Turkish migrants in Denmark and Germany. In fact, half of the respondents in Germany did not obtain more than lower secondary education, and only 37 percent received higher secondary or tertiary education. However, whilst people in the lowest educational category constitute the biggest individual group in Denmark, too, taken together, 49 percent of the Turkish respondents in this country received higher secondary or tertiary education. Turkish respondents in Italy particularly stand out given that the distribution of educational titles amongst them resembles those observed for Romanian migrants, with the biggest group having received tertiary education, and a clear majority holding at least a higher secondary degree.

The relative share of those who received secondary or tertiary education varies between 83 and 88 percent for Romanian respondents in the three countries under observation. Furthermore, even in Germany, where the highest share of Romanian migrants in the lowest educational category is situated, this group did not exceed four percent of the total sample. In contrast to Turkish migrants in Italy, Romanians in the same country possess the lowest average educational level, when compared to their compatriots in the other two countries. However, these differences are

107

Sociodemographic description not that profound and are clustered towards the upper end of the scale due to the comparatively high share of individuals holding higher secondary degrees.

Approximately one quarter of each of the Turkish samples received their highest educational degree in their current CoR. A comparable distribution can also be observed in the Romanian sample in Denmark, with lower percentages observed in Italy and Germany. This corresponds to the information on the age of migration presented above, which indicated that members of all the Turkish samples and Romanians in Denmark migrated on average at an earlier age than Romanians in Germany and Italy. Consequently, respondents in the latter two samples were more likely to have already finished their schooling or university education before migrating. Therefore, they were less likely to seek additional educational titles in the CoR. Accordingly, and unsurprisingly, the analysis shows a highly significant negative relation between higher age at migration and the probability of possessing an educational title from the CoR (table not included). Finally, when comparing formal educational outcomes, in all samples, the relative share of respondents who reached tertiary education is considerably higher for those who received their highest educational title in the CoR. Furthermore, the average level of education is in all but one sample higher for those educated in the CoR. Only in the case of Romanians in Germany is the average at the same level for those educated in either country.

While higher education does not guarantee labour market participation, educational capital certainly enhances its probability. The labour market status of an individual, in turn, is not only an important aspect of structural integration but can also influence other dimensions of integration. Regular employment can, for example, provide the financial means to participate in social and cultural events and offer possibilities to interact with other individuals on a regular basis. Figure 3 provides detailed information regarding the labour market integration of the migrants in all six subsamples.

108

Sociodemographic description

Figure 3: Labour market participation (percent)

0 10 20 30 40 50 60 70 80 90 100

Working

Non-working

Retired

Unemployed

Other

TU in DK TU in DE TU in IT RO in DK RO in DE RO in IT EUCROSS, 2016. N = 1,502

The data show that the majority of respondents in all samples participate in the labour market. However, the percentages are lower for Turkish migrants than for Romanians (between 54 and 61 percent, and between 68 and 70 percent respectively). In regard to those who do not have paid work but do not consider themselves unemployed either, for example, because they are in education, the differences between both nationalities are less pronounced. Approximately one in ten respondents of the Turkish sample in Denmark and of the Romanian sample in Germany falls into this category, while it is one in five in the remaining four samples. However, there are clear gender differences regarding the working and non-working status (detailed data not included in Figure 3). While three quarters of all male Turkish respondents have a paid job, only 40 percent of women do. At the same time, every third Turkish woman stated that she is not working, but only three percent of men did so. These differences are also visible to a lesser extent in the Romanian samples. While 61 percent of the Romanian women and 77 percent of the men have a paid position, every fifth woman and almost every tenth man fall into the non-working category. The percentage of women who explicitly stated that they are doing housework or caring for children is highest amongst Turkish migrants in Germany (39 percent). Approximately, one in

109

Sociodemographic description five Turkish and Romanian women in Italy and Turkish women in Denmark are also in this category. In contrast to their Turkish counterparts, a significantly smaller share of Romanian female respondents in Germany (12 percent) and Denmark (7 percent) could be described as housewives. Unsurprisingly, the share of retirees is highest for the two samples in Germany, which also have the highest percentages of respondents who are 60 years or older.47 Between 5 to 15 percent of the respondents in the individual samples are unemployed, with Turkish migrants being more strongly affected than Romanians in all CoRs. The difference in relative unemployment levels between both nationalities is, with four percentage points, the smallest in Germany. This is compared to eight percentage points in Denmark and nine percentage points in Italy. For migrants from both surveyed countries, unemployment is highest in Denmark, followed by Italy and then Germany. Taken together, the employment and unemployment rates indicate a higher labour market integration of intra-EU migrants compared to third country migrants. This is in line with similar findings reported by authors who compared the corresponding values of migrants from different countries in Denmark (Pedersen 2005, 71-72) and Germany (Schunck 2014, 192).48

The financial situation of the respondents is likely to be strongly related to their labour market participation and, therefore, constitutes the next aspect to be examined. With regard to the following values it should be kept in mind that they do not directly refer to personal income, but rather to the general financial situation of the households (see chapter 3.4). Consequently, these values can also be influenced by the income of respondents’ partners. The data show that only a small minority of the migrants from both countries find themselves in a difficult or very difficult financial situation (11 percent of Turkish and 5 percent of Romanians respondents). At the same time, only the Romanians show a clear majority (63 percent) for the statement that the financial means available to their households allow for a comfortable or very comfortable life. Less than one third consider their situation neither particularly difficult nor comfortable. Amongst Turkish migrants, the group of respondents who live (very) comfortably is only slightly bigger than the number of respondents in economically less well-off households which make ends meet (46 versus 42 percent).

47 22 percent of Romanians and 21 percent of Turks in Germany fall in this age category. 48 Regarding the figures presented by Schunck, this statement refers to the values of countries which are explicitly mentioned by the author and not included in his broader ‘other’ categories or summarised as Ex-Yugoslavia.

110

Sociodemographic description

Figure 4: Household financial situation49

0 10 20 30 40 50 60 70 80 90 100

TU in DK 2 8 38 38 14

TU in DE 2 11 45 32 9 Find it very difficult Find it difficult TU in IT 1 9 44 38 6 Make ends meet

RO in DK 2 26 56 15 Living comfortably Living very comfortably RO in DE 3 29 56 11

RO in IT 2 9 36 49 3

EUCROSS, 2016. N = 1,502.

Figure 4 shows the distribution of respondents across the five categories describing the household financial situation by country. From this it is apparent that migrants of both nationalities were on average financially best off in Denmark. Amongst the Turkish migrants, households in Germany were described as being least wealthy overall. In the case of Romanian migrants, those in Italy described their households as least wealthy. However, even in this sample, the share of individuals who live in a comfortable or very comfortable economic household situation is with 52 percent just as high as in the Turkish sample which is on average the best off (i.e., Turks in Denmark).

However, this overall picture changes somewhat if another indicator of wealth is considered, specifically the question of whether the respondents or their partners owned real estate. In Figure 5 the absolute number of property owners is combined with the indicator of household financial situation. Since not all respondents own real estate, the five categories contain considerably fewer cases. Therefore, each of the two categories that signalled a comparatively comfortable or uncomfortable household situation are combined. The displayed figures include property in the country of residence, the country of origin and third countries.

49 Figures might not sum up to 100 percent due to rounding and missing values.

111

Sociodemographic description

Figure 5: Real estate ownership by household financial situation (absolute numbers)

0 20 40 60 80 100 120 140 160

TU in DK 6 31 65

TU in DE 14 49 57 (Very) difficult TU in IT 10 39 59 Make ends meet RO in DK 3 14 89 (Very) comfortable RO in DE 1 37 97

RO in IT 17 58 92

EUCROSS, 2016.

Interestingly, for both nationalities, the same subsamples which were, on average, the least satisfied with their financial situation, are those for whom real estate ownership is most common. This holds true for the Romanians in Italy just as much as for Turkish respondents in Germany. In both cases, the combined number of individuals who see their households as being in a difficult or only satisfactory situation, but nevertheless owned property, was higher than in the other two samples of each nationality. Furthermore, this is not just the case if all instances of property ownership are pooled, but also if ownership in the CoO and the CoR are examined separately (data not presented). Figure 5 also clearly shows that property ownership is by no means limited to those who describe themselves as living comfortably. On the contrary, in the Turkish sample in Germany property ownership is even more common amongst those households who did not report a high level of financial comfort, i.e., that make ends meet at best. Even though this is not the case for Romanians in Italy, still, of all the Romanian samples this one has, in absolute and relative terms, the highest share of property owners who do not believe their households to be in a financially comfortable situation.

These findings underline the need to include the question of property ownership in the assessment of structural integration in the used data. This is particularly important in light of the fact that, for survey methodological reasons, no direct measurement of income was used. In this context, it should be kept in mind that property does not necessarily create income, but when used personally for housing or business, it might actually be conceived as an additional cost

112

Sociodemographic description factor in daily life. For instance, costs could be associated with loans, mortgages, taxation, or the maintenance of holdings. Accordingly, property could in some cases contribute to a more negative assessment of net household wealth. Such an effect would be most likely to materialise in low-income households. In a nutshell, this means, that the overall relation between property and household income might be positive, i.e., comparatively better-off households are more likely to own real estate than poorer ones. However, on the other hand, it is conceivable that some respondents who own real estate would assess their household’s financial situation more positively if they did not own property requiring regular monetary investments.

Focusing on real estate ownership in the CoR, the data (table not included) show that comparable numbers of respondents in four out of the six samples hold such property (between 21 and 28 percent). Outstanding are Turkish migrants in Denmark who show the lowest rate with 14 percent, and Romanians in Italy who have the highest rate with 41 percent. If the rates are compared on a CoR level, in both Denmark and Italy the share of Romanian respondents who own property is nearly double the percentage amongst Turkish migrants. Only in Germany are the ownership rates more comparable at 28 percent for Romanians and 23 percent for Turks. This means that the probability of real estate ownership in the settlement context does not only differ when migrants in one CoR are compared to those in another, but also between the two migrant groups within each country. However, this is not surprising if the above-reported differences in age, duration of settlement, education and labour market participation are taken into account.

4.4 Summary This study investigates migrants of Turkish and Romanian nationality who, on a general level, constitute the biggest migrant groups in the European Union. Historical migration patterns reported in the literature for these migrants are also visible in the data. For example, in Turkish samples in Denmark and Germany work-related migration motives were more often mentioned by earlier migration cohorts, while family reunification gains importance amongst later ones. This corresponds to findings about the changing character of migration to both countries from ‘guest worker’ to family-oriented mobility in the last three decades of the 20th century. Whereas the Turkish respondents reported a longer duration of stay than the Romanians in both countries, the contrary holds true for the two samples in Italy. The median age of most samples lies between

113

Sociodemographic description

40 and 47 years. Only Turkish respondents in Italy and Romanians in Denmark are somewhat younger and on average in their early 30s. The data show a rather balanced distribution of gender.

Generally speaking, the surveyed Romanian migrants have, on average, a higher level of education than the Turkish migrants. While higher secondary and tertiary education is prevalent amongst the Romanian respondents in all countries, the Turkish samples show a higher degree of variation. Turkish respondents in Germany are, on average, the least educated: approximately half of them do not have more than lower secondary education. In contrast, one fifth of the surveyed Turkish migrants in Italy completed higher secondary education and four out of ten even hold a college or university degree. Unsurprisingly, the two Romanian samples that show considerably lower levels of CoR education compared to the other samples are also those in which respondents tended to migrate at a higher age, making it more likely that they had completed their education before engaging in international migration. Finally, it is shown that the reported average level of education is in most samples higher amongst CoR than CoO educated respondents.

The majority of respondents in all samples participates in the labour market. Nevertheless, it can be observed that the rates of the Turkish migrants are, with 54 to 61 percent, lower than those of Romanian respondents, of whom approximately 70 percent have a regular occupation. Furthermore, the gender gap in labour market participation is more pronounced amongst Turkish than Romanian respondents. The two samples with the lowest average educational level (Turks in Denmark and Germany) also show the lowest participation in the labour market.

Finally, the analysis demonstrated that Romanian respondents, on average, not only reported higher levels of education and participation in the labour market, but also gave more positive assessments of their households’ current financial situation. While migrants from both countries described their households as best off in Denmark, Turkish respondents struggle financially the most in Germany, and Romanians in Italy. At the same time, the analysis of property ownership revealed that even households which saw themselves in less favourable financial situations are often in possession of this type of non-monetary capital. These findings support the decision to include this aspect in the analysis as an additional indicator of structural integration.

114

5 The transnationalism of Turkish and Romanian migrants

This chapter examines the degree of transnationalism present within the studied samples. Transnationalism, as discussed above, is understood as a multidimensional phenomenon. The respective systematisation is the focus of the first part of this chapter. Here, the empirical connection between the indicators within each of the dimensions and subdimensions of transnationalism will be shown. Building on this theoretical and empirical foundation, the discussion will then turn to the extent to which the respondents possess transnational human capital, their involvement in transnational networks, and the degree to which they engage in transnational practices. The inclusion of a diverse range of indicators will allow drawing a detailed picture of respondents’ cross-border engagement. Besides the discussion of these aspects on the level of the individual samples, the comparison of the two migrant populations in all three CoRs is of special interest. The chapter will, furthermore, take the geographic range of transnationalism into consideration. This is a key contribution to the literature, as within empirical migration research transnational engagement has hitherto almost exclusively been examined as a range of activities connecting the countries migrants live in with those they originated from. More specifically, this chapter will go beyond this dichotomy by discussing in detail the geographic dispersion of respondents’ transnational networks and the range of their cross-border mobility. Building on these descriptive results, a final analytical subchapter will investigate factors which facilitate or impede transnational practices. The chapter closes with a summary of its key findings.

5.1 Multidimensionality of transnationalism To analyse transnationalism, this study applies a modified version of the systematisation of Kuhn (2011, 814). The details regarding the operationalisation are provided in chapter 3.4 and build directly on the theoretical arguments put forward in the second chapter. In brief, the study will differentiate between the three principal dimensions of transnational human capital, transnational background and networks, and transnational practices. Within the latter dimension the subcategories of communication, economic activities, mobility, political engagement and consumption will be differentiated from each other. In the following, it shall be tested whether the relations which are assumed – based on theoretical considerations – to exist between indicators within each dimension or subdimension can be shown empirically. Hence, this analysis

115

The transnationalism of Turkish and Romanian migrants serves to confirm the used operationalisation. The method employed to this end is multiple correspondence analysis (MCA, see chapter 3.2). This technique allows displaying the data included in the rows and columns of contingency tables graphically within a two-dimensional space. The logic is that the relative positions which indicators take in relation to each other within this space show whether they are empirically related. The closer that indicators are situated together, the stronger their connection (Blasius 2010, 367).

Figure 6 shows separate MCA plots for the indicators in the dimensions of transnational human capital, transnational background and networks, and for the different subdimensions of transnational activities. Regarding the latter, only political transnationalism is not included in this analysis as the data only include a single indicator for this phenomenon (see chapter 3.4). All indicators are coded in an ascending manner, i.e. zero means that the respective aspect of transnationalism is not present. In dichotomous variables, the value one represents observations which exhibit the transnational trait. In non-dichotomous variables, higher values indicate different degrees of the occurrence of the particular aspect of transnationalism. If the indicators in a dimension are indeed related to each other, similar values should cluster close together. Most importantly, zero values should form clusters that can be clearly differentiated from higher values indicating the occurrence of transnationalism. Table 6 provides an overview of all indicators used in Figure 6.

116

The transnationalism of Turkish and Romanian migrants

Figure 6: MCA plots of indicators within the dimensions and subdimensions of transnationalism

Transnational human capital Transnational background and networks

.2 .5 .1 nc2 oc0 oc2 nb2

0 me0 me1 al1 oc1 al0 na2

nc0 0

-.1 nb0 na0 dimension 2 ( ( 0.8%) 2 dimension dimension 2 (13.3%) 2 dimension nc1 po1po0 nb1

na1

-.2

-.3 -.5 -.4 -.3 -.2 -.1 0 .1 .2 -.5 0 .5 1 dimension 1 (88.6%) dimension 1 (69.3%) coordinates in principal normalization coordinates in principal normalization N = 1,486 N = 1,405

Transnational communication Economic transnationalism

1 .1

cm4 .05 cs4 .5 cp4 re0

0 re1pa1 pb0 pa0 cp0 rm0 cs0 rm1 cm3 cm0 0 cp3 cs3 -.05

cm1 cs1 cp1 -.1

cp2 ( 1.7%) 2 dimension dimension 2 (20.0%) 2 dimension

-.5 pb1 cm2

cs2 -.15

-1 -.2 -1 -.5 0 .5 1 -.1 -.05 0 .05 .1 .15 .2 dimension 1 (61.9%) dimension 1 (76.0%) coordinates in principal normalization coordinates in principal normalization N = 1,490 N = 1,420

Transnational mobility Transnational consumption

.4 .6

.4 .2

ta0 .2 tb3 tb0 ms3

0 ta3 tv4 ms0 ta1 sh1 tb1 0 tb2 ms1 sh0 tv0

ta2 ms2 tv1 dimension 2 ( ( 3.1%) 2 dimension dimension 2 ( ( 4.1%) 2 dimension tv3

tv2

-.2

-.4 -.4 -.4 -.2 0 .2 .4 -.6 -.4 -.2 0 .2 .4 dimension 1 (95.8%) dimension 1 (92.6%) coordinates in principal normalization coordinates in principal normalization N = 1,488 N = 1,421

117

The transnationalism of Turkish and Romanian migrants

Table 6: Indicators used in the multiple correspondence analysis of transnationalism (Figure 6) Dimensions Subdimensions Indicators Labels Values At least moderate knowledge al 0 = no; 1 = yes of an additional language Transnational Familiarity with other oc 0 = no; 1 = one; 2 = two or more human capital countries Additional migration me 0 = no; 1 = yes experience Partner who lives in another po 0 = no; 1 = yes country Family/friends from CoO in na 0 = none; 1 = a few; 2 = a lot Transnational CoO background Family/friends from CoO in nb 0 = none; 1 = a few; 2 = a lot and networks 3rd country Family/friends from 3rd nc 0 = none; 1 = a few; 2 = a lot country living abroad Via phone and VoIP cp 0 = never; 1 = less often than once a month; 2 = at least once a month; 3 = at least once a week; 4 = every day Via mail and email cm 0 = never; 1 = less often than Transnational once a month; 2 = at least once communication a month; 3 = at least once a week; 4 = every day Via social networking sites cs 0 = never; 1 = less often than once a month; 2 = at least once a month; 3 = at least once a week; 4 = every day Sends at least once a year re 0 = no; 1 = yes money abroad Transnational Economic Received money from abroad rm 0 = no; 1 = yes practices transnationalism (last 12 months) Property in CoO pa 0 = no; 1 = yes Property in a third country pb 0 = no; 1 = yes Number of trips to CoO ta 0 = none; 1 = one; Transnational (last 24 months) 2 = two; 3 = three or more mobility Number of trips to 3rd tb 0 = none; 1 = one; countries (last 24 months) 2 = two; 3 = three or more Percentage of messages from ms 0 = 0%; 1 = 1-32%; abroad (excluding spam) 2 = 33-65%; 3 = 66-100% Watches foreign-language TV tv 0 = never; 1 = less often than Transnational content once a month; 2 = at least once consumption a month; 3 = at least once a week; 4 = every day Shopping abroad sh 0 = no; 1 = yes (last 12 months)

With the exception of economic transnationalism, all MCA plots displayed in Figure 6 show the same basic geometric pattern, which confirms a general relationship between the indicators included in the respective dimensions and subdimensions. Indicators that signal the absence of a specific type of transnationalism (zero values) are located in relative proximity to each other near

118

The transnationalism of Turkish and Romanian migrants the right end of the x-axis. The position of these indicators is distinct from those that signal various levels of transnationalism and that are positioned to their left.

The first plot of Figure 6 contains the measurement of transnational human capital, including indicators of at least moderate knowledge in a third language (al1), familiarity with two or more other countries besides CoO and CoR (oc2), and of additional migration experience (me1). These elements are all located in relative proximity to each other in the negative region of the x-axis. However, knowledge of a third language and familiarity with two or more additional countries are more closely related to each other than to additional migration experience. Furthermore, the plot indicates that familiarity with only one other country (oc1) is relatively independent of the other indicators as it is closer to the cluster indicating the absence rather than the presence of language knowledge and migration experience (al0, oc0, me0).

Besides the clear pattern along the x-axis, the indicators of transnational background and networks (second plot in Figure 6) also show a distinct order on the y-axis which is defined by the relative size of networks. The networks in question consist of friends and family from the CoO who are living in the CoO and third countries, and of individuals of third country origin in any country other than the CoR. Indicators of a smaller number of contacts of all three types (na1, nb1, nc1) cluster in the negative quadrant while the indicators of larger number of contacts (na2, nb2, nc2) are situated in the positive range of the y-axis. This means that respondents with only a few contacts of any type, have more in common with each other than with respondents who have a lot of contacts in any of the categories, and vice versa. The fact of having a partner abroad (po1) is situated in-between. However, this indicator is comparatively close to the origin, meaning that it does not have a strong relation with the other aspects.

The first type of transnational activities, that is transnational communication (third plot in Figure 6), also shows the above-described pattern on the x-axis with higher communication frequencies at one end and the absence of communication on the other. The extreme points of the y-axis are defined by those individuals who communicate on a daily basis across borders (cp4, cm4, cs4; upper left quadrant) and those who do so at least once a month but less than every week (cp2, cm2, cs2; lower left quadrant). Apparently, the differences are less pronounced between individuals who engage in transnational communication less than once a month (cp1, cm1, cs1)

119

The transnationalism of Turkish and Romanian migrants and those who do so on a weekly basis (cp3, cs3, cm3), as shown by the relative proximity of these indicators to each other in-between the two discussed clusters.

The MCA plot of transnational mobility (fifth plot in Figure 6) shows an increase of trips from right to left on the x-axis. Furthermore, a high association between three or more trips to either the CoO (ta3) or a third country (tb3) can be observed. The same holds true for one visit to a third country (tb1) and two trips to the CoO (ta2). The indicator for respondents who travelled to third countries twice during the last 24 months (tb2) is situated between those groups, while individuals who visited their CoO only once during the same time (ta1) are more similar to those who did not leave the CoR at all (ta0, tb0).

In the case of transnational consumption (sixth plot in Figure 6), besides the common pattern across the x-axis, the examination of the distribution across the y-axis shows that those indicators which characterise a stronger form of transnationalism are situated in the upper quadrant. These aspects are shopping abroad (sh1), daily consumption of TV content in a third language (tv4), and foreign origin of at least 66 percent of all received messages via email or social networking sites (ms3). At the same time, indicators of lower frequencies of TV consumption and lower percentages of messages from abroad (tv1-3, ms1-2) are grouped together in the lower quadrant.

As mentioned above, the MCA plot of one subdimension of transnational activities, namely of economic transnationalism (fourth plot in Figure 6), deviates from the geometric pattern along the x-axis. Four indicators are included in this dimension: having property in the CoO, having property in a third country, sending money abroad at least once a year (i.e., remittances), and having received money from abroad during the last 12 months. The MCA plot shows a strong relation between remitting money (re1) and the existence of property in the CoO (pa1). Both indicators are situated on the negative side of the x-axis together with those individuals who did neither receive money (rm0), nor own property in a third country (pb0). However, the latter indicator is situated very close to the origin and does, therefore, relate very little to all others. This already implies that individuals who received money (rm1) or have property in a third country (pb1) are situated in the positive space defined by the x-axis. Hence, the two groups of indicators constitute opposite poles on this axis. While all four phenomena clearly have economic implications, this means that they are indicative of two distinct forms of economic transnationalism. As the four indicators are undoubtedly connected by the fact that they involve

120

The transnationalism of Turkish and Romanian migrants border-crossing connections to real estate and financial transfers this should not be understood as a cause to refuse the subdimension as such, but rather to take its somewhat more diverse character into account during the remainder of the analysis. Furthermore, all zero values (re0, rm0, pa0, pb0) are situated on the positive side of the y-axis while all substantial indicators (re1, rm1, pa1, pb1) are on the negative side. However, this aspect is of minor importance since the distance between most indicators is minimal on this axis.

In conclusion, the presented MCA results show that the individual indicators included in the dimensions and subdimensions are empirically related to each other. In all cases, indicators signalling the absence of transnationalism are situated close to each other, as are indicators signalling its presence. Furthermore, the displayed patterns are logical in the sense that indicators of more frequent engagement in different transnational practices cluster together within a given dimension. Consequently, these results support the proposed multidimensional operationalisation of transnationalism and the theory-driven allocation of indicators to the specific dimensions and subdimensions. The following part of the chapter will discuss in detail the degree to which the different aspects of transnationalism can be observed in the analysed data.

5.2 Transnational human capital The first aspect of transnationalism which shall be discussed is the transnational human capital of respondents. Table 7 presents an overview regarding the three indicators used to measure this dimension across all six samples.

Table 7: Features of transnational human capital (percent)

Turkish migrants in … Romanian migrants in … Denmark Germany Italy Denmark Germany Italy

At least moderate knowledge of a 3rd language 33 21 58 91 62 62 Familiarity with 3rd countries (one or more) 80 70 68 70 61 59 Additional migration experience 13 4 30 25 16 11 EUCROSS, 2016. N =1,502

As argued in earlier chapters, additional language knowledge is an important skill with respect to transnational activities as it facilitates inter-cultural communication, though not necessarily guaranteeing its success. The present data show that 93 percent of all Romanian and 57 percent of all Turkish migrants had previously learned another language. However, the mere fact of having learned a language at some point does not necessarily mean that these individuals can actually

121

The transnationalism of Turkish and Romanian migrants converse in it. Therefore, the first row of Table 7 shows the percentage of respondents of each sample who stated that they have at least moderate knowledge of an additional language. It is not surprising that the samples which show, on average, a higher level of education (see Table 5), also consist of higher percentages of individuals who speak a third language. This assessment holds true not only if both nationalities are compared but also for the comparison of the individual samples in all three countries. Generally speaking, a clear majority of respondents in all Romanian samples possess this kind of human capital. While the same is only true for Turkish migrants in Italy, it is nevertheless remarkable that – given the reported average level of formal education – still one in five Turkish respondents in Germany can converse in a third language.

The second row of Table 7 shows the degree to which respondents in the various samples are familiar with countries besides their CoO and CoR. As can be seen, at least six out of ten respondents in all samples are familiar or very familiar with at least one third country. Contrary to language knowledge, which is a rather specific form of human capital, general knowledge of another country is more common amongst the Turkish than the Romanian migrants. The comparison of the figures in the first two rows also suggests that both forms of transnational human capital are not necessarily directly related, as language knowledge is by far less common amongst Turkish migrants than familiarity with other countries.

The third aspect of transnational human capital captured in the data is migration experience. As argued above, this is seen as an indirect indicator for the acquisition of skills and cultural or social knowledge relevant in transnational settings. Such knowledge might pertain, for example, to traditions or social conventions in other countries. As this work focuses on migrants, all respondents have by definition acquired migration experience. Accordingly, the percentages presented in the last row of Table 7 only refer to additional stays of three or more consecutive months in a country other than the respondents’ CoO and CoR. These values are, by a large margin, the lowest amongst all three indicators. Three of the six samples stand out: First of all, Turkish migrants in Germany of whom only four percent have ever lived in a third country. Second, Turks in Italy and Romanians in Denmark of whom 30 percent and 25 percent, respectively, have additional migration experience. Hence, while the level of additional migration experience is comparable for Turks in Denmark, and for Romanians in Germany and Italy, the three aforementioned samples demonstrate considerably higher or lower percentages. These differences can partly be explained by general migration patterns. Germany, which hosts the

122

The transnationalism of Turkish and Romanian migrants biggest Turkish community and Italy, for which the same can be said regarding Romanians, are also the two countries which the highest share of respondents of either nationality migrated directly to. On the contrary, Turkish migration to Italy and Romanian migration to Denmark are more recent phenomena. Therefore, it does not come as a surprise that for the respondents in these samples, both countries were more often destinations of onward migration. Indeed, additional analyses (table not included) show that those European countries in which approximately ten percent of the two samples’ respondents with prior migration experience had lived before are countries that host comparatively large communities of migrants from either CoO. Hence, these are countries which are commonly thought of as destinations of Turkish or Romanian migration.50

5.3 Transnational background and networks Transnational background constitutes the second dimension of transnationalism. Just as transnational human capital, it is not defined by specific practices but might enhance the respondents’ probability to engage in such. The importance that cross-border social relations play in the emergence and endurance of transnational spaces has been stressed by various authors (e.g., Glick Schiller, Basch, and Szanton Blanc 1999, 85-88; Mau 2007, 94-95).

The first row of Table 8 shows that most migrants from both CoOs have people abroad whom they know well (friends) or are directly related to (family). The percentages vary between 72 and 90 percent which is quite similar to the findings of Fauser et al. (2015, 1502), according to which 80 percent of migrant respondents in the German SOEP sample have contacts in other countries. However, keeping in mind that all respondents of the survey at hand were actually born abroad it is surprising that in three samples approximately one quarter of the participants does not have any friends or family members in other countries than their CoR. This is particularly true in the case of Turkish migrants in Germany of whom 28 percent did not report such contacts.

50 For Turkish respondents in Italy, these countries are France, Germany, the Netherlands, Switzerland, and the United Kingdom. Taken together 52 of the 76 respondents (68 percent) have lived in one of those countries before migrating to Italy. For Romanian respondents in Denmark, these countries are Germany, Italy, Spain and the United Kingdom. Taken together 36 of the 63 respondents (57 percent) have lived in one of those countries before migrating to Denmark.

123

The transnationalism of Turkish and Romanian migrants

Table 8: Contacts abroad (percent)

Turkish migrants in … Romanian migrants in … Denmark Germany Italy Denmark Germany Italy

Any family members, in-laws or friends who live outside the CoR 90 72 90 88 78 75 Partner lives outside the CoR 3 2 8 5 4 3 EUCROSS, 2016. N = 1,502.

Table 8 further shows that the contacts respondents have in other countries only include their life partners in a small minority of instances. Unfortunately, this information is rarely reported in other studies complicating the comparison of these results with earlier findings. One of the exceptions is the study of immigrants and their direct descendants in Norway by Carling and Hoelscher. The authors analyse data which had been collected amongst migrants from ten different countries, finding that three percent of the cumulative sample reported having a partner abroad (Carling and Hoelscher 2013, 943). However, their sample did include Turkish respondents and, unfortunately, they do not report these values on a CoO basis. In the present data, the cumulated percentages of respondents with life partners in other countries vary between three percent in Germany and five percent in Italy. Hence, these values are on a comparable level to those found in Norway. Nevertheless, such a comparison should only be seen as an approximation due to the very different composition of these samples and the methodological differences between this study and the one presented by Carling and Hoelscher.

In order to allow a more nuanced understanding of personal networks abroad, it is important to understand the diversity of migrants’ personal transnational networks. Fortunately, the data used here allow for the differentiation of these networks regarding the origin and the location of contacts (Figures 7 to 9).

124

The transnationalism of Turkish and Romanian migrants

Figure 7: Family, in-laws, and friends from the CoO living in the CoO (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 74 14

TU in DE 58 12 A lot TU in IT 52 34 A few RO in DK 52 32

RO in DE 31 40

RO in IT 34 36

Figure 8: Family, in-laws, and friends from the CoO living in a third country (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 26 45

TU in DE 14 29 A lot TU in IT 32 46 A few RO in DK 26 46

RO in DE 17 43

RO in IT 18 40

Figure 9: Family, in-laws, and friends from a third country not living in the CoR (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 9 20

TU in DE 6 12 A lot TU in IT 15 32 A few RO in DK 5 26

RO in DE 8 21

RO in IT 3 16

Figures 7-9: EUCROSS, 2016. N = 1,502. Taken together, Figures 7 to 9 show that border-transcending social networks of migrants from both nationalities centre on individuals with CoO background. Seventy percent or more of respondents in all samples have friends or family who are of CoO origin and also live in their CoO. If differentiated by CoR, 87 percent of the migrants in Denmark, 79 percent in Italy, and 70 percent in Germany have co-ethnic family or friends in their CoO. Several studies from the

125

The transnationalism of Turkish and Romanian migrants

Netherlands do not directly report the percentage of respondents who have similar contacts but mention the percentage of their sample which communicated with family members in the CoO. This, in turn, presupposes the existence of corresponding contacts. The cumulative values of 87 percent (Bilgili 2014a, 292), 88 percent (Snel, Engbersen, and Leerkes 2006, 292), and 92 percent (Schans 2009, 1174) for all migrants in these studies indicate that the existence of family members in the CoO is at least as frequent amongst them as amongst the participants of this study.

Furthermore, in five out of six samples, more than half of the respondents have co-ethnic contacts living in third countries (Figure 8). These are individuals who are themselves Turkish or Romanian migrants, or who descended from them. Out of all six samples, Turkish migrants in Germany are the least likely to have this kind of connections to other countries. In the light of Turkish migration history, and taking into account that Germany hosts the biggest population of Turkish origin in the European Union, this finding is not surprising. Similarly, Romanian respondents in Italy and Germany, which have been major destination countries for mobile Romanians since the downfall of socialism, are less likely to be in contact with compatriots in third countries than their co-nationals in Denmark. When considering more dense networks (i.e., having a lot of contacts), it can be observed that the values of migrants in main destination countries (Turks in Germany; Romanians in Germany and Italy) cluster between 14 and 18 percent, while the values of the remaining samples are at considerably higher levels with 26 to 32 percent. This corresponds to the trend which could be expected in view of existing migration statistics. In a nutshell, if approximately 1.6 million Turkish nationals live in Germany, while eighteen thousand reside in Italy, then the probability that a randomly chosen Turkish citizen will know a co-national in Germany is higher than the chance that he or she will know such a person in Italy. Consequently, a Turkish respondent in Italy is more likely to know another Turk in Germany than the other way around.

The comparison of Figures 7 to 9 shows that individuals of CoO background are clearly most prominent in the respondents’ networks abroad. However, it is clear that these networks are not limited to this group alone. Figure 9 reveals that even amongst Turkish migrants in Germany, and Romanians in Italy (the two groups with the lowest values for this measure) approximately one in five individuals has friends or family members of a different origin living abroad.

126

The transnationalism of Turkish and Romanian migrants

Generally speaking, the majority of respondents in five out of the six samples has friends or family in third countries, i.e., countries other than the CoO and CoR (figure not included). Only amongst Turks in Germany, the proportion of the sample that reported such connections constituted with 45 percent a minority. The Romanian sample in Italy has the second lowest value but still a clear majority of 60 percent of respondents has family or friends in third countries. The same holds true for 63 percent of the Romanians in Germany, for 74 percent of respondents of both nationalities in Denmark, and for 81 percent of the Turkish sample in Italy. These findings indicate that the cross-border relations of most of the surveyed migrants are not limited to a CoR-CoO dichotomy.

For a more detailed analysis, respondents were asked to report the specific countries in which they had friends and family members, thus indicating the individual scope of cross-border social connections. For the purpose of this analysis, these values were accumulated on a sample level and visualised with the help of maps. These maps can be seen in Figure 10 and show the relative size of each sample that has personal connections to a specific country. To allow for a meaningful comparison, not all countries to which connections were mentioned are highlighted. Instead, this is the case only for those countries that were named by five or more percent of the respondents in a sample. The darker a country, the higher the percentage of respondents who reported having family members or friends living there.51

51 The corresponding percentages are presented in Appendix 4.

127

The transnationalism of Turkish and Romanian migrants

Figure 10: Geographic scope of transnational networks Map 1: Turkish migrants in Denmark Map A: Romanian migrants in Denmark

Map 2: Turkish migrants in Germany Map B: Romanian migrants in Germany

Map 3: Turkish migrants in Italy Map C: Romanian migrants in Italy

Percentage of a sample that has friends or family in a country Country of residence

5-10 % 21-30 % 41-50 % 61-70 % 81-90 % Turkish migrants 11-20 % 31-40 % 51-60 % 71-80 % 91-100 % Romanian migrants

128

The transnationalism of Turkish and Romanian migrants

The maps in Figure 10 show that cross-border networks of a relevant part of all samples go beyond the duality of the country of birth and country of residence. This also holds true for Turks in Germany of whom nearly a quarter has close contacts in the Netherlands and France. Mau and Mewes show that the geographic structure of transnational networks of German nationals is highly related to historical, political, social, and cultural relations between Germany and other countries. Furthermore, they stress the importance of Germany’s history of emigration and immigration in this regard (Mau and Mewes 2007, 213-15; Mau 2007, 99-101). A somewhat similar conclusion can be drawn from the present data.

As discussed, research on transnational networks which link migrants with different countries, i.e., which goes beyond the CoR-CoO dichotomy, is scarce. Therefore, no direct frame of reference for the observed geographic distribution of contacts is available. That said, if the connections displayed in Maps 1 to 3 are seen in the light of broader Turkish migration history, in line with the arguments of Mau and Mewes, a concentration of contacts in Western Europe is obvious. In fact, comparatively high shares of all Turkish migrants have close contacts in Germany, the Netherlands, France, and Austria. These countries have also been the major destination countries of Turkish migration since the middle of the 20th century (Karakaşoğlu 2008, 1054; Oğuz 2012,15). To a lesser extent, the same is true for Belgium and Switzerland. Hence, it seems that the structure of personal networks is heavily influenced by the migration system which started to unfold with the signing of contract worker agreements between Turkey and a number of Northern European countries in the late 1960s.52 This seems especially likely in the case of Turks in Denmark and Germany, as communities in both countries show high rates of early work-related migration, which later gave way to a stronger role of family reunification. In both instances, a considerable part of the sample belonged to the age cohort which embarked on comparatively highly regulated work migration targeted at a variety of European countries. It is reasonable to assume that some members of this cohort who ended up working in different European countries have relations to each other which pre-date their migration. Such relations might in the following have been strengthened by the relative geographic proximity of their CoRs, and the resulting possibilities for private, family, and business-related exchanges. Icduygu and Sirkeci (1998, 7) identify four consecutive periods of Turkish work-related migration between the early 1960s and late 1990s. According to these authors, who based their analysis on

52 For brief discussions of the concept of migration systems see Fawcett (1989) and Kritz and Zlotnik (1992).

129

The transnationalism of Turkish and Romanian migrants data collected by the Turkish Employment Service, labour migration was focussed on Europe in the first period (1961-1973/74), while it later shifted towards Arab countries, to a lesser extent to Australia and finally to Eastern Europe. The fact that Turkish respondents in none of the samples reported a large number of connections to either of these regions indicates the influences of a distinct migration system. This system developed between Turkey and European countries which recruited contract workers, and influenced the flow of migratory movements and cross-border networks of these individuals. The influence of these connections between Western Europe and Turkey is also apparent in the geographic distribution of contacts within the networks of Turks in Italy. Germany, which hosts the biggest Turkish community outside Turkey itself, is, again unsurprisingly, the most prominent third country to which networks of Turkish migrants in Denmark and Italy extent. While nearly nine out of ten Turks in both countries have connections to Turkey, more than six out of ten also have family or friends in Germany. The fact that these figures are nearly identical for Turkish migrants in both CoRs underlines that these connections are not determined by geographic proximity alone.

Regarding the spatial structure of Romanian migrants’ networks depicted in Maps A to C of Figure 10, a similar effect of general migration patterns can be seen. Even though labour oriented emigration from Romania only reached significant levels after the end of socialism, several destinations have been very prominent at different times since then. With regard to Europe, this pattern has already been described above (chapter 4.1). Germany, Italy and Spain are identified as the main destination countries in the literature. Less pronounced, but still significant immigration of Romanians is reported for Austria, France and the United Kingdom (Sandu 2005, 50; Ban 2012, 132; Andrén and Roman 2016, 254). This structure is also visible in the networks of Romanians shown in Figure 10, with Germany, Italy, and Spain being the most cited locations of family and friends in third countries. Other European states were mentioned less often, corresponding to their secondary position as destination countries.

A good proportion of all Romanian samples also reported personal connections to the United States and Canada. The literature shows that pronounced migratory flows developed between Romania and both countries during the late 1990s (Sandu 2005, 39; Andrén and Roman 2016, 256), i.e., coinciding with the growth in Romanian migration towards Italy and Spain. The overlap of networks between the migratory flows to Western Europe and North America stands in contrast to the above-discussed absence of such crossovers in the networks of Turkish

130

The transnationalism of Turkish and Romanian migrants respondents with co-nationals who were part of migratory flows to different world regions. This can partly be explained by the fact that Turkish migration in its infancy was highly formalised and structured by intergovernmental agreements which fostered the establishment of relatively separate migration systems between the CoO and different world regions. In the Romanian case, such institutional interventions were not only absent but the focus of migratory flows also changed to different countries after a comparatively shorter period. This made it more likely that individuals with prior connections to each other migrated not only to different countries but also to different continents. Amongst Turkish migrants, only those in Italy reported a noteworthy degree of connections to the United States.

The fact that only the samples in Denmark show significant connections to Sweden and (in the case of Turkish migrants) to Norway, suggests that cultural and political relations between the CoR and other countries also have a potential influence on migrants’ personal networks.

5.4 Transnational practices The following subsections will shed light on different types of transnational activities. These are aspects which are typically in the focus of transnational migration research and constitute – in their entirety – the third dimension of transnationalism in the systematisation used here.

5.4.1 Transnational communication Cross-border communication is a transnational activity of high importance as it allows individuals who reside in different countries to be and to remain part of each other’s social lives. Furthermore, such communication has the potential to stabilise transnational networks and, thereby, transnational lifestyles over time. Almost all respondents who reported having friends or family abroad also communicated with them during the 12 months preceding the survey. In fact, only 18 respondents (i.e., 1.2 percent of the overall sample) stated that they have such contacts abroad but did not communicate with them in the specified period. Overall, this means that between 71 and 88 percent of the respondents in all samples communicated with friends and family abroad during the 12 months preceding the survey.

In addition to the mere occurrence of communication, its frequency is also of interest as it can be seen as an indicator of the density of the transnational connections. To analyse the use of the different means of communication (phone/VoIP, mail/email and SNS) in more detail, the frequency of communication (ranging between every day and never) stated by the respondents for

131

The transnationalism of Turkish and Romanian migrants each type of communication was reverse coded from zero to four. Figure 11 displays the mean values of all six samples excluding respondents who have no friends or family abroad. In addition to the three mentioned ways of communication, a cumulative variable (Communication (all means)) was added, which recorded the highest frequency in which respondents contacted their friends and family members abroad using any of the three methods of communication. Hence, this variable is an approximation of the general frequency of communication. However, it is still likely to underestimate the actual contact rate as it only depicts one (the most frequently used) means of communication and not the combined use of all three.

Figure 11: Use of different means of transnational communication (means)53

3.5

3.0

2.5

2.0

1.5

1.0

0.5 TU in DK TU in DE TU in IT RO in DK RO in DE RO in IT

Communication (all means) Phone or VoIP Mail or e-mail Social networking sites

EUCROSS, 2016. N = 1,223.

Starting at the top of Figure 11, it becomes apparent that in all samples those respondents who have friends and family in other countries are – on average – in contact with them at least once a week using either verbal communication, mail, email, or SNS. The differences are marginal with Romanians having 0.2 to 0.3 higher mean values in each country. This is an important finding as it indicates that transnational networks indeed constitute a social space which allows individuals in different countries to participate regularly in each other’s lives. This interpretation is strengthened by the fact that the respondents also have a clear preference for verbal communication which makes it more likely that they indeed engage in direct conversations with their counterparts. The first two graphs also demonstrate that, on the sample level, the quantity

53 Scale: 0 - never; 1 - Less often than once a month; 2 - At least once a month; 3 - At least once a week; 4 - Every day.

132

The transnationalism of Turkish and Romanian migrants and quality of contacts are not entirely congruent, since those samples where a smaller percentage of respondents reported having friends or family abroad do not differ notably from the other samples in terms of average communication frequency. This means having fewer friends and family members abroad does not automatically lead to less transnational communication.

The lower three graphs in Figure 11 clearly show that respondents in all samples, on average, prefer direct verbal exchanges on a transnational scale over non-verbal forms of communication. The average Romanian and Turkish migrant in all three countries tends to have this kind of contact abroad at least once a week. Differences are again visible but marginal. Turks in Denmark have the lowest mean value. However, this mean value of 2.6 still indicates that they are more likely to have weekly rather than monthly calls. The high frequency suggests that talking to people in other countries is an everyday activity for the surveyed migrants. Since it is unlikely that respondents have dramatically good or bad news to report each week, it is safe to assume in accordance with Pries (2008, 49) that for most of them long-distance calls have lost their main character as emergency signals or short signs of life, which may have been their function in earlier periods. In fact, messages sent via social networking sites, which are the second most frequently used method of international communication, are more likely to be used for this purpose nowadays (while not being limited to this function). In particular, the availability of cheap international telephone connections is an aspect which stabilises transnational social relations over time, as it allows the participating parties to stay directly connected and involved in each other’s daily life (Vertovec 2004, 220). Emails and letters, which are usually much longer than the aforementioned messages and substantially less direct than telephone calls, are only sent infrequently by the average respondent. The mean values of Turks in Germany and Denmark, and Romanians in Denmark cluster around the less than once a month category, while Turks in Italy have the strongest tendency towards monthly contacts of this kind. Romanians in Germany and Italy are in-between. These findings support the assessment that contemporary transnationalism sets itself apart from historical cross-border connections, among other things, by offering the possibility of immediate communication across large distances that allows people to be part of each other’s daily life, in spite of considerable geographic distances between them (Faist, Fauser, and Reisenauer 2013, 54).

133

The transnationalism of Turkish and Romanian migrants

5.4.2 Economic transnationalism The following section concentrates on monetary exchanges and property abroad as indicators of economic transnationalism. Figure 12 shows that a significant proportion of respondents in all samples send money abroad. While the Turkish sample in Germany has the smallest percentage of respondents who do so, the share amongst Turks in Italy is only a few percentage points higher. Nevertheless, it is apparent from these figures that more than a quarter of respondents in both samples remit money. The same holds true for 38 percent of the Turks in Denmark and between 33 and 42 percent of the Romanian samples. Leaving aside their CoR, Romanian respondents are more likely to send remittances (37 percent) than Turkish participants (31 percent). The comparison of these figures with other studies is complicated, as they differ in the employed measurement instrument.

Figure 12: Respondents sending money abroad (percent)54

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100

TU in DK 9 23 5 0.4

TU in DE 2 21 3 0.8

TU in IT 8 13 6

RO in DK 12 15 8 0.8

RO in DE 14 19 9

RO in IT 10 20 2 0.4

At least once a month At least once a year Less than once a year Frequency unknown

EUCROSS, 2016. N = 1,502. First, there are a number of studies which did not use a specific time frame when inquiring remitting behaviour. This means respondents were asked whether they regularly send remittances or whether they ever sent remittances in general. Therefore, the data of such studies correspond to the above-discussed values. Studies of this type tend to report comparatively high remittance rates: for Latin-American migrants in the United States reported rates are at 74 (Itzigsohn and

54 Figure 12 is based on the answers to two questions: The first one asked whether respondents did ever send money to other countries for other reasons than to purchase goods or services. Only those who answered affirmatively were consequently asked how often they did so. The category Frequency unknown refers to those respondents who confirmed that they were sending money abroad, but did not indicate how often.

134

The transnationalism of Turkish and Romanian migrants

Giorguli Saucedo 2002, 776), approximately 55 (Castañeda, Morales, and Ochoa 2014, 320), and 47 percent (Waldinger 2008, 26) of the respective samples. Regarding European countries, Castañeda et al. (2014, 320) reported that 56 percent of the Northern-African respondents they surveyed in France remitted money. The results of Snel et al. (2006, 292) showed that 33 percent of the respondents in their survey sent remittances. The latter study collected data in the Netherlands and analysed the most diverse sample of all mentioned surveys in terms of migrants’ origin. However, with individuals from Japan and the United States, this sample includes migrants from two countries which differ substantially in their economic power from the other four considered regions of origin (i.e., Morocco, Dutch Antilles, Iraq, former Yugoslavia), as well as from the two CoOs of the respondents in this analysis. Unsurprisingly, respondents from the United States and Japan were much less likely to remit money. If both subpopulations are disregarded, the share of respondents in the remaining sample who sent remittances rises to 48 percent.55 All aforementioned percentages cumulate respondents of different nationalities within a given CoR. The corresponding values resulting from the data used in this study are 37 percent (Denmark), 35 percent (Germany) and 31 percent (Italy). These values are much lower than those reported for migrants in the United States and France. The same holds true with regard to the findings pertaining to migrants in the Netherlands if respondents from the United States and Japan are disregarded.

A second group of studies employed slightly more differentiated measurements of remitting behaviour, allowing their authors to provide information on the proportions of the respective samples that sent money to their CoO within different time frames (e.g., on a monthly or yearly basis). The following discussion will use yearly remitting behaviour as a frame of reference. If cumulative values are used, the results from the data at hand (i.e., 30 percent in Denmark, 28 percent in Germany and 26 percent in Italy) are comparable to the value of approximately 32 percent of respondents who reported sending remittances at least once a year in the Dutch survey by Dekker and Siegel (2013, 7). The mean value resulting from the figures of the very diverse sample which Carling and Hoelscher (2013, 949) realised in Norway is, however, again considerably higher with approximately 45 percent of all respondents remitting on a yearly basis.56 At the same time, the results of these authors also underline that remitting behaviour

55 The mentioned percentages were calculated by the author based on information included in Table 1 in Snel et al. (2006, 292). 56 Own calculation based on the provided values for individual samples (Carling and Hoelscher 2013, 949).

135

The transnationalism of Turkish and Romanian migrants differs heavily depending on the CoO, with values ranging from 25 percent (Iran) to 65 percent (Somalia) of the corresponding subsamples. Of the Turkish respondents in their study, 39 percent remit money at least once per year, which is a somewhat higher percentage than that of the Turks in each of the three CoOs surveyed here. However, they also report that 6 percent of the Turks surveyed in their study send money to the CoO at least once a month, which is comparable to the values reported in Figure 12. Taken together, it can first be concluded that the remittance rates reported in the literature vary strongly, both within and across countries. Second, the rates in the present data are on the lower end of the range of previously reported results and are closer to the values observed in other European countries than to the values reported for migrants in the United States.57

The results presented in Figure 12 stress that the remitting of money is not a majority phenomenon amongst the surveyed migrants. Yet, the minority that does engage in this cross- border activity is considerable with between one in five and one in three migrants sending money at least once a year. It is also important to note that the frequency in which funds are remitted does not provide any indication of the financial commitment’s value. Several amounts sent every week might, for example, still be less in total than a larger sum sent once a year. Furthermore, it has to be taken into account that the used categories are constructed as continuums, i.e., the range of answers could, therefore, cover everything from at least once a month or a year, to every day, or in 11 out of 12 months, respectively.

A last noteworthy feature of the remitting behaviour of respondents in this study is that the numerically most important beneficiaries of these activities are family members to whom between 91 percent of all Romanians who engage in this activity and 72 percent of their Turkish counterparts send money. Non-family members are, on the contrary, seldom beneficiaries of such transfers (Romanians: 9 percent; Turks: 32 percent).

While the data also show that some of the migrants were actually the recipients of remittances, this occurred to a much smaller degree. Amongst Turkish migrants in Denmark and Germany, the values are negligible, with less than four percent of the respondents in either case receiving

57 Schunck (2014, 187) states that 13 percent of the migrant respondents in the German SOEP remitted money on a yearly basis. However, his measurement deviates strongly from all of the above-mentioned: First, he reports the average percentage of individual measurements between 1996 and 2009 and, second, the question used to collect this data referred explicitly to the year before the respective survey and was not of a general nature (i.e., referring to remittances on a yearly basis).

136

The transnationalism of Turkish and Romanian migrants money from abroad. The values of Romanians in Germany and Italy are also low at seven percent each. These low values are similar to the figures reported by Dekker and Siegel (2013, 19) for migrants from four different countries in the Netherlands, which ranged from approximately two to nine percent. However, the share of individuals who receive money from abroad is significant, both in the case of Romanians in Denmark (18 percent) and amongst Turkish migrants in Italy (27 percent). The data show that transnational financial relations are mostly sustained amongst close relatives, since individuals of this group are most often mentioned as those who send money to their migrated family members, too.

Contrary to many other studies, the investigation of economic transnationalism was not limited to monetary transfers in the used survey. In addition, respondents were asked whether they or their partner, if they indicated that they had one, owned real estate in the CoO or a third country. As could be expected based on the overall figures presented in the demography chapter above, property ownership in the CoO is not a majority phenomenon amongst the surveyed migrants. However, there is a noticeable minority of approximately one third of the respondents in all three Turkish samples, who own real estate in Turkey.58 The values vary much more amongst Romanian migrants, with 26 percent in Denmark, 40 percent in Germany, and 46 percent in Italy. It appears that Germany is the CoR in which the difference in CoO property ownership between both nationalities is smallest (approximately five percentage points), which corresponds to the reported findings on property ownership in the CoR. Compared to figures presented in the literature, these percentages are relatively high, with Carling and Hoelscher (2013, 943) and Snel et al. (2006, 292) reporting that approximately 20 percent of their respondents in Norway and the Netherlands held property in the CoO.59 This value corresponds to the 22 percent that Hiebert and Ley reported for immigrants from various backgrounds in Vancouver, Canada (Hiebert and Ley 2006, 76). The figures presented by Snel and co-authors, however, also underline the large variation present in property ownership between migrants of different nationalities within the same CoR. Finally, Landolt points out that 39 percent of the migrants from El Salvador in the CIEP study owned property in their CoO (Landolt 2001, 224), which is a higher percentage than that of all Turkish samples and of Romanians in Denmark, but relatively comparable to the overall value amongst Romanians in Germany.

58 31 percent of Turkish respondents in Italy, 32 percent in Denmark, and 35 percent of those living in Germany own real estate in Turkey. 59 At CoR level the cumulative values of migrants from both CoOs seen together are: 29 percent in Denmark and 38 percent in Germany and Italy.

137

The transnationalism of Turkish and Romanian migrants

Regarding the country in which property is located, there is a clear pattern amongst those who own real estate: 56 percent of all Turkish owners and 41 percent of the Romanians have only real estate in the CoO but not the CoR. On the contrary, 25 respectively 31 percent of the respondents have property in the CoR but not in the CoO. The data do not include information regarding the nature of the property, i.e., whether it is a residential building or used for commercial activities, nor whether it is used by the respondents, their partner, or rented out to third parties. Nevertheless, these figures mean that it is less common for migrants to own property in the CoR, which they could potentially permanently occupy, than to have real estate in their CoO, which is either used by third parties or vacant most of the year. There are a number of possible explanations for this tendency. Just to mention a few, the probability of inheriting real estate in the CoO has to be considered higher than in the CoR, given that the respondents migrated to the latter, which means that their family origins and possible family property are situated in the former. Furthermore, it has been pointed out in the literature that owning a house in the origin community does not only mean that migrants have a vacation home to come back to but also that such a residence can be a status symbol and sign of their social position in time of physical absence (Pauli 2008, 179; Erdal 2012, 636). Additionally, real estate is often considered a solid investment and a means of securing wealth independently of economic and political developments. From this point of view, it might be simpler and more cost effective to buy property or build a house in the CoO, due for instance to smaller costs of the required resources and labour or less complicated building codes in comparison to the CoR. Finally, in the case of Turkish migrants, residential status could also be an issue as none of the respondents have been naturalised in their respective CoR. Hence, property in the CoO might also serve as a ‘safety net’ (Erdal) in the sense that it provides migrants with a place to go back to if they were not able to stay in the CoR (Erdal 2012, 632). It should also be mentioned that smaller but still considerable portions of the Turkish (18 percent) and Romanian property owners (27 percent) have real estate in both countries. On the contrary, property ownership in a third country is negligible, with only 12 Romanian migrants and four Turkish respondents reporting ownership of such real estate.

5.4.3 Transnational mobility As previously discussed, mobility is an important aspect of transnationalism since it is not only a cross-border activity in its own right but also has the potential to strengthen or establish transnational networks. Figure 13 shows that the absolute majority of all samples have been

138

The transnationalism of Turkish and Romanian migrants internationally mobile during the 24 months preceding the survey. The first bar indicates the percentage of each sample which undertook at least one trip to a country other than the CoR. For both nationalities, this percentage is the lowest in Germany. However, even in this country 86 percent of Turkish and 87 percent of Romanian respondents had been abroad during the preceding two years. Amongst the Turkish samples, individuals in Denmark were most likely to have undertaken such trips. In the Romanian case, individuals in Denmark and those in Italy showed a similarly high percentage of trips abroad. Romanian migrants were generally more likely to have engaged in cross-border mobility.

From a CoR perspective, differences between both nationalities are smallest (ranging between one and three percentage points) for the overall mobility. On the contrary, the biggest differences are visible for trips to countries other than the respondents’ CoO (third bar). While approximately one third of Turks in Germany visited a third country during the 24 month period, the same holds true for nearly two thirds of the Romanian respondents in this country. In the case of visits to the CoO, both samples show a much smaller difference of only six percentage points. In Denmark, the distinction between the samples is less pronounced but the share of respondents who have visited a third country is still 17 percentage points higher for Romanians than for Turks. In the case of visits to the CoO, both samples are only four percentage points apart.

139

The transnationalism of Turkish and Romanian migrants

Figure 13: At least one trip abroad during the last 24 months (percent)

0 20 40 60 80 100

93 TU in DK 88 52

86 TU in DE 83 32

89 Any trips abroad (total) TU in IT 76 67 Trips to the CoO 92 Trips to a 3rd country RO in DK 84 69

87 RO in DE 77 63

92 RO in IT 87 56

EUCROSS, 2016. N = 1,502. Figure 14: Three or more trips abroad during the last 24 months (percent)

0 20 40 60 80 100

29 TU in DK 14 20

25 TU in DE 19 10 Three or more trips abroad 52 TU in IT 40 28 Three or more trips to the CoO 59 RO in DK 39 Three or more trips to a 42 3rd country

53 RO in DE 34 35

42 RO in IT 31 21

EUCROSS, 2016. N = 1,502.

140

The transnationalism of Turkish and Romanian migrants

Only in Italy, the destination country for trips abroad does not seem to strongly influence the relative difference between the two nationalities, which is at 11 percentage points for visits to the CoO and also to a third country (Figure 13). Generally speaking, the values indicate that the degree of transnational mobility might be more strongly influenced by differences between migrant groups when it is directed to countries other than to the CoO. It can also be observed that Turkish respondents living in Denmark and Germany visited their CoO more often than Romanians in both countries, while the contrary is true with regard to visits to third countries. Only in Italy were visits to the CoO more common for Romanians, and visits to third countries for Turkish respondents. While Figure 13 shows on a general level that visits to the CoO were most common, it also indicates that a considerable number of individuals in all samples visited other countries too. With the exception of Turks in Germany, this was even the case for the majority of respondents in all groups.

As stated, there are unfortunately no published results to which the mentioned figures on transnational mobility can be compared to as a whole. This is due to the fact that migration research mostly concentrates on visits to the CoO as far as transnational mobility is concerned. However, even the presented values pertaining to this phenomenon cannot be compared directly with the literature. The main reason lies in the different ways of measuring mobility. Whereas in this study respondents were asked for specific experiences during the 24 months preceding the survey, others asked in a more general way how often respondents would return to their CoO (Dekker and Siegel 2013, 8), whether they visited the CoO once a year (Guarnizo, Portes, and Haller 2003, 1226), or whether migrants had visited the CoO at all since they arrived in the CoR (Waldinger 2008, 14). Finally, in one project the measurement instruments referred to periods of different lengths in the individual waves of a longitudinal study by reporting the total number of visits to the CoO since the last data collection point (O'Flaherty, Skrbis, and Tranter 2007, 826). In addition to these methodological differences, there are also a number of important features directly related to the surveyed populations and the countries in which the studies were realised, which could impact comparisons (e.g., cultural dissimilarities between target populations, differing legal statuses and frameworks or sociodemographic profiles of migrant populations).

Of the available information, Schunck’s findings seem the most comparable to the present data, as the question in the SOEP, which he uses in his analysis, also refers to a two-year period and as the data pertain to migrants in Germany, i.e., one of the countries under investigation here.

141

The transnationalism of Turkish and Romanian migrants

Schunck reports that 69 percent of the migrants in his sample had visited their CoO (Schunck 2014, 184). The cumulative values in the present study are at 80 percent in Germany, 81 percent in Italy, and 86 percent in Denmark. While these figures are comparable, the data at hand suggest slightly higher rates. However, it should also be taken into account that the sample analysed by Schunck was far more diverse than this one in terms of included CoOs (see Schunck 2014, 184). This potentially influenced the mobility behaviour-related results. The above-mentioned longitudinal study design used by O’Flaherty et al. also provides one possible comparator as one of the time frames that their question referred to also covered a 24-month period. The authors state that approximately 37 percent of all respondents had visited their CoO during this time (O'Flaherty, Skrbis, and Tranter 2007, 828). Finally, using the CIEP data, Guarnizo et al. report an average value of approximately 19 percent for migrants who visited their CoO on an annual basis (Guarnizo, Portes, and Haller 2003, 1226). In turn, the data used here show that between 54 and 62 percent of migrants from all three countries visited the CoO twice or more often during the 24-month period, which can be seen as a somewhat crude approximation of an indicator for undertaking such trips on a yearly basis. Based on these comparisons, it can be concluded that the respondents visited their CoO on average at least as often as the migrants analysed in other studies. There are in fact strong indications that they did so somewhat more often than immigrants residing in the United States and Australia. This, however, does not come as a surprise considering the respective distances and associated costs. It can be assumed that in most cases it is easier to visit Turkey and Romania from either of the three CoRs, than for migrants in the United States or Australia to visit their CoO. Migrants travelling from Denmark, Germany, and Italy can, for example, use private cars and commercial bus services to visit Turkey or Romania (in combination with ferries depending on their exact place of residence). However, such relatively cheap modes of transportation are either not available in the case of Australia and the United States, or their use would take a considerable amount of time due to the larger geographical distances. In the same sense, it can be assumed that the regular costs of air travel between the countries in this study are lower than for connections between Australia or the United States and the origin countries of the immigrants surveyed in the above-mentioned studies. In fact, several authors have made similar observations (e.g., Morawska 2002, 140; Bilgili 2014a, 298).

142

The transnationalism of Turkish and Romanian migrants

Going a step further than Figure 13, i.e., beyond the simple occurrence of cross-border mobility, Figure 14 shows the percentages of respondents from all samples who undertook three or more trips abroad during the 24-month period preceding the survey. This means that these individuals travelled abroad, on average, more than once a year. Again, this is not at all an uncommon occurrence as the lowest value (Turks in Germany) still means that at least one in four respondents in this sample is engaged in this form of frequent transnational mobility. Once more, this phenomenon is generally more common amongst Romanian as opposed to Turkish respondents. However, Turks in Italy constitute the one sample of this nationality in which more than half of the respondents undertook at least three trips abroad. On the contrary, Romanians in Italy are the only sample from this CoO of which less than 50 percent did this.

If the values for all three CoRs are compared within each national group, a clear pattern emerges amongst Romanians: for all three indicators in Figure 14 (i.e., three or more trips to any country, to the CoO and to third countries), the sample in Denmark shows the highest values while the one in Italy shows the lowest, with Germany in-between. Regarding Turkish migrants, the field is slightly less structured. The Italian sample has the highest values for all three indicators, while the Danish sample scores second highest on overall mobility and visits to third countries, but also contains, in comparison, the smallest share of respondents who visited the CoO three or more times. This leads to the interesting observation that for both nationalities in Denmark, frequent visits to third countries were more common than frequent visits to the CoO. It can be assumed that this comparatively high occurrence of trips to third countries is at least partially due to the country’s size and its geographic position: Denmark is a comparatively small country which at the same time is connected by land, bridges and short-haul ferries to its neighbouring countries. This means that the latter can be reached quickly and without the necessity of an unduly high investment of time or money from any point in the country. This also means that the costs of visiting a neighbouring third country are, in most cases, likely to be significantly smaller for Turkish and Romanian migrants than a visit to their CoO, as the latter would necessarily comprise of a much longer journey by car, bus, or train if not a flight.

On the other hand, the geographic position and shape of the Italian mainland peninsula have the consequence that the accessibility of neighbouring countries is more dependent on the specific location of settlement within the CoR. Hence, a migrant in Southern Italy might be just as likely to depend on long-distance public transport or planes in order to visit either the CoO or a third

143

The transnationalism of Turkish and Romanian migrants country. Consequently, these respondents might prefer to visit their CoO instead of a third country if comparable investments in time and money have to be realised for both.

Figure 15 shows the main areas of transnational mobility of the respondents during the 24 months preceding the survey. Again, the maps in Figure 15 show only those countries which were mentioned by at least five percent of the any given sample.60 Thus, it should be kept in mind, that some of the respondents have visited additional countries, just as there have been sporadic contacts with other countries than those depicted in Figure 10, i.e., additional countries and regions that have been mentioned by less than five percent of a given sample.

60 The corresponding percentages are presented in Appendix 5.

144

The transnationalism of Turkish and Romanian migrants

Figure 15: Geographic scope of transnational mobility Map 4: Turkish migrants in Denmark Map D: Romanian migrants in Denmark

Map 5: Turkish migrants in Germany Map E: Romanian migrants in Germany

Map 6: Turkish migrants in Italy Map F: Romanian migrants in Italy

Percentage of a sample that visited a country Country of residence

5-10 % 21-30 % 41-50 % 61-70 % 81-90 % Turkish migrants 11-20 % 31-40 % 51-60 % 71-80 % 91-100 % Romanian migrants

145

The transnationalism of Turkish and Romanian migrants

Like in the case of the personal networks, the maps in Figure 15 show a concentration of cross- border mobility in Europe. More specifically, it is even more Europe-centric than the reported transnational networks, since no non-European destination was mentioned by at least five percent of any group. It can be assumed that this is also due to the higher costs associated with visits to countries such as the United States or Canada. The pattern regarding the diversity of destinations follows the one that has been described above for highly frequent mobility (three trips or more within 24 months). This means a higher number of countries was visited by at least five percent of the Romanians in Denmark than by their compatriots in Germany. Romanians in Germany, however, visited a higher number of countries than Romanians in Italy. This replication pattern is also visible in the case of Turkish migrants, amongst which those in Italy mentioned the most countries, followed by their compatriots in Denmark and Germany. Hence, the relation between those two aspects indicates that Turkish and Romanian migrants who undertake several trips abroad do not necessarily visit the same destinations every time. Furthermore, it shows a diversity of migrants’ transnational mobility which would not be captured by the sole concentration on visits to the CoO.

The mobility patterns of Turkish migrants in Germany are highly concentrated on neighbouring countries. The two destinations most often mentioned by Turks in Denmark, i.e., Germany and Sweden, share direct borders with their CoR as well. With the Netherlands and Norway, two additional countries which are in close proximity to Denmark are in the larger group of countries mentioned less often. Besides the high value of Germany, Turks in Italy also often mentioned countries which bordered their CoR (i.e., France, Switzerland and Austria). Particularly for Romanians in Denmark and Germany, a direct geographic connection between the CoR and destination country seems to be less important. Respondents in both samples frequently mentioned destinations located at a longer distance from the country that they were living in (e.g., Spain, Italy and France in the case of Romanians in Denmark; Italy and Spain in the case of Romanians in Germany).

Overall, this means that amongst the most frequently mentioned destination countries of transnational mobility are often, but not exclusively, those which are close and can, therefore, quickly be reached, which makes them attractive destinations for short visits. Furthermore, shorter distances usually also mean that less monetary resources have to be invested in such trips than if individuals travelled to more distant destinations.

146

The transnationalism of Turkish and Romanian migrants

Besides geographic proximity, the comparison of the maps in Figures 10 and 15 show that there is a connection between the geographic scope of social networks and transnational mobility. Similar to the way in which individual networks concentrate on countries with large communities of Turkish and Romanian migrants, mobility patterns do. The congruence of migrants’ space of mobility and the geographic scope of their transnational networks is most pronounced for the Turkish samples. Only amongst Turkish respondents in Italy, more than five percent mentioned a specific destination (Greece) in which less than five percent of the sample has family or friends. Again, this picture is somewhat different for Romanian migrants: especially respondents in Denmark and Germany cited a number of countries in which no significant part of these samples has friends or family members, as destinations of cross-border mobility. The visible, and at least partial, congruence between the geographic patterns of networks and mobility is nevertheless unsurprising. While close contacts in a certain location enhance the probability of visits, (repeated) visits to a location foster the formation or consolidation of personal relations with individuals in these locations.

Interestingly, between ten and twelve percent of Romanian respondents in all three CoRs mentioned Hungary in the context of mobility; however, the five percent threshold was not reached in any of the samples for this country as a location of friends and family. This means that the transnational mobility of Romanians is, in contrast to their social networks, not merely focussed on Western Europe. Given the geographic position of Hungary, however, it is possible that visits to it had the character of transits on the way to Romania itself. The same might be true for the Czech Republic, which was mentioned by Romanian migrants in Denmark, given that the two main North-South highways connecting Germany to Central- and South-Eastern Europe cross Germany’s border in Saxony and Bavaria. While, in the former case, travellers to Romania would be likely to continue through the Czech Republic, a small strip of Slovakia and Hungary, in the latter case travellers would have to transit Austria and Hungary before reaching Romanian soil.61 Transit might also be the reason why Greece was mentioned as a destination by eight percent of the Turkish migrants in Italy, as ferries (e.g., those embarking in the Italian city of Brindisi) connect the country to Greece and would allow travellers from Middle and Southern

61 78 out 83 Romanians who visited Hungary and 22 of 23 who visited the Czech Republic also mentioned at least one visit to Romania.

147

The transnationalism of Turkish and Romanian migrants

Italy to reach Turkey by car much more easily compared to the driving time62 and distance coming from Denmark or Germany.63

While visits to friends and family are important reasons for cross-border mobility, it stands to reason that there are other motivations too. In his research on the general population of Germany, Mau showed that international trips were mainly undertaken as part of vacations (mentioned by 61 percent of the respondents in his survey). Other reasons were much less prominent, e.g.: visiting friends and family at 13 percent, professional reasons at 12 percent and other private reasons at 10 percent (Mau 2007, 135). Similar to Mau’s survey design, respondents who visited third countries were asked to state the main reasons for their trips.64 The two most important reasons for transnational mobility amongst the German population identified by Mau are also most prominent amongst the migrants analysed here. Taken together, 66 percent of all Turkish, and 81 percent of all Romanian respondents who visited a third country did so in order to go on vacation. Visiting friends or family was mentioned as a main reason by 44 percent (Turks) and 24 percent (Romanians) of these migrants. The somewhat higher significance of such visits compared to the German population is not surprising as Mau’s respondents also reported significantly lower numbers of family members and friends abroad (Mau 2007, 104, 106). With respect to the individual samples, Turkish migrants in Denmark are an exception as they are the only sample in which slightly more respondents said they travelled to visit friends and family (63 percent) than to go on vacation (60 percent). In all other samples, the above-described hierarchy is visible with between 69 and 84 percent of the mobile respondents giving vacations, and between 19 and 35 percent citing visits as motivation for transnational mobility. Professional reasons for trips were most often cited by Turks in Italy, with 21 percent of the mobile respondents. For four of the other five samples, these values lay between 9 and 13 percent, which is similar to the 12 percent reported by Mau. Turkish respondents in Germany were least likely to visit another country for professional purposes (4 percent).

62 Excluding time spent on the ferry. 63 All Turkish migrants in Italy who visited Greece also mentioned at least one visit to Turkey. As mentioned in chapter 3.4, the respective item asked the respondents to think only about trips which included at least one overnight stay. At first view, this seems to contradict the stated interpretation that visits to some of the countries might mainly have had the character of transits. However, the item which included this restriction referred to trips as a whole and not to visits to specific countries. Hence, it is reasonable to assume that respondents not only mentioned their final destination but also other countries which they might have passed through. 64 This information is not available for each individual country that has been visited but was collected as a follow-up to the question which asked respondents to state the countries they had visited in the previous 24 months.

148

The transnationalism of Turkish and Romanian migrants

5.4.4 Political transnationalism If respondents are compared by nationality, it becomes apparent that the percentages amongst the potential transnational voters who actually participated in the last general election in their CoO are nearly identical, with 14.5 percent for Turkish and 14.2 percent for Romanian respondents. This is surprising in so far that the possibilities for casting a vote were quite different for emigrants of both nationalities at the time of the survey. While Romania already has had a procedure in place by 2008 in which diplomatic missions could set up polling stations in foreign countries (Popescu 2012, 100), a similar policy was only implemented by Turkey in 2014. Until the necessary legal changes, which occurred in 2012, Turkish citizens abroad could only cast their vote at polling stations at the Turkish border (Okyay 2015, 119, 143). This means, despite the fact that voting was logistically more difficult for them – by a very small margin – more Turks than Romanians made use of this possibility. In this context, it should be mentioned that voting is mandatory in Turkey (Spierings 2014) but not in Romania (IDEA 2017). However, the fine Turkish non-voters face amounts to less than 10 € (Ataman and Ghelli 2014). Furthermore, it is not clear whether this applied to emigrants at the time of the survey, as mechanisms for absentee voting were only established thereafter (see above). In general, it is unlikely that respondents travelled to their CoO in order to avoid such a minor punishment for non- compliance. However, it is possible that socialisation in a system of compulsory voting had a positive effect on Turkish electoral participation. Seen at a CoR level, engagement in the measured form of political transnationalism was highest in Italy (19 percent), followed by Denmark (13 percent), and Germany (12 percent). While both samples in Italy were equally likely to vote, in Denmark the Romanian and in Germany the Turkish respondents were more likely to do so than the survey participants of the respective other nationality.

The comparison of the reported values with those of other studies can only serve as an indicator of a general tendency, as the likelihood to engage in the process of absentee voting is likely to be influenced by a number of aspects which vary strongly between surveyed populations and countries. Most prominent amongst these are institutional factors of both the CoO and the CoR, aspects of political culture and education that differ between CoOs, and even within a given cohort, by the specific election itself, and the media coverage it receives. Nevertheless, the reported values are comparable to the 15 percent of CoO electoral participation reported by Waldinger (2008, 26) across different migrant groups in the United States.

149

The transnationalism of Turkish and Romanian migrants

5.4.5 Transnational consumption The last type of transnational practice focusses on activities of an everyday character. The first observation which can be made in this regard is that approximately every third Romanian and every sixth Turkish migrant engaged in cross-border shopping during the 12 months prior to the survey without being physically mobile. As can be seen in Figure 16, this activity was most common amongst Romanians in Denmark, of whom nearly 60 percent engaged in it, while less than 20 percent of Turks in this country did so. A quarter of Turkish migrants in Italy bought something abroad, which makes them the subsample of this nationality for which this activity was most common. For migrants from both countries, buying abroad was most unusual when they lived in Germany, even though differences between the two samples are clearly visible, with Romanians being much more likely to order something abroad than Turks. In the latter case, less than six percent of the sample bought something abroad. Overall, only amongst Romanians in Denmark a majority of all respondents engaged in cross-border shopping, while this activity has to be seen as a clear minority phenomenon in all other samples.

Nevertheless, interesting similarities emerge if the locations of the providers of the bought goods and services are considered. If the samples are aggregated by CoO65, it becomes clear that for virtual shopping of migrants of both nationalities, Germany, the United Kingdom, and the United States were most important. This is the case, even though migrants in Germany could – for logical reasons – not state this country, which means that the United Kingdom and the United States had a higher chance of being mentioned in the overall sample. Nevertheless, Germany outranked the other two countries quite clearly in both cases: slightly more than half of all Romanian migrants66 who bought something abroad, and 36 percent of Turkish migrants67 who did the same, purchased goods or services from providers in Germany. Unfortunately, no data are available regarding the content of those transactions, but the high share of providers frequented in Germany could be directly related to the low figures regarding trans-border purchases by migrants in this country (Figure 16). It is, in fact, possible that migrants in Germany bought from the same providers in their CoR as their respective compatriots in the other two countries. However, if they did so, those transactions would not have been captured by the used survey instrument. Furthermore, it is worth mentioning that cross-border shopping did not primarily take

65 N for Turkish migrants = 119; N for Romanian migrants = 239. No information on a CoR level is provided due to the small number of cases. 66 UK: 28 percent, USA: 20 percent. 67 UK: 13 percent, USA: 19 percent.

150

The transnationalism of Turkish and Romanian migrants place between the countries of residence and origin. Only seven percent of the Romanians who engaged in cross-border shopping and 26 percent of their Turkish counterparts purchased something from providers in their CoO.

Figure 16: Features of transnational consumption (percent)

100

0 TUTU in DK TU TUin DE TU TUin IT RO ROin DK RO inRO DE RO inRO IT in in in IT in in in IT DK DE DK DE Foreign-language TV consumption Electronic messages from abroad Shoping abroad Weekly/daily Less frequent 66-100 % 33-65 % 1-32 %

EUCROSS, 2016. N = 1,502

Moving on to the consumption of foreign-language TV content as the second indicator in this dimension, Figure 16 shows that this habit was more widespread amongst Romanian, rather than Turkish respondents in Denmark and Germany. This is not only true for the general tendency to consume such content but also when the frequency of this activity is taken into account. This means in both countries respondents in those samples which show higher average levels of education and foreign language capacities (compare Table 5 and Table 7) tend to engage more often in this type of media usage. However, in Italy, where differences in education and foreign language knowledge are less strong between Romanian and Turkish respondents, this pattern cannot be observed. On the contrary, in this case, Turkish migrants have a stronger tendency to watch foreign language content, even though they are, on average, slightly less educated and less

151

The transnationalism of Turkish and Romanian migrants likely to possess skills in a third language. Figure 16 also shows that Romanians in Denmark are more likely to watch foreign language content than the respondents in all other samples, while Turkish respondents in Denmark did so more frequently than Turks in Germany, as well as Romanians in Germany and Italy. It is very likely that this effect is related to the availability of foreign language content through traditional TV channels. Unlike in Germany and Italy where most foreign language content is dubbed, and, consequently, original versions are mostly available through pay TV or the internet, shows and movies are usually broadcasted in the original language including subtitles in Denmark (Safar et al. 2011, 6-7).

Not counting spam, 60 percent of all Romanian and 47 percent of all Turkish respondents received electronic messages from outside their CoR during the 12 months preceding the survey. However, providing some more detail, Figure 16 shows that in three of the six samples (Turks in Denmark and Germany, and Romanians in Italy) only a minority of respondents received messages from abroad. Furthermore, of the three categories which signal that respondents did at least receive some message from other countries, the lowest one (i.e., 1-32 percent) contains in all samples more cases than either of the other two categories. Thus it becomes apparent that overall, most respondents did receive no or less than one third of all their messages from other countries. Consequently, such messages do not dominate this form of media consumption. Nevertheless, in three samples (Turks in Italy; Romanians in Denmark and Germany) more than one quarter of the respondents received one third or more of their message from another country.

5.5 Determinants of transnational practices The preceding sections provided detailed information on the degree to which Turkish and Romanian migrants in the three countries under observation engage in transnational activities. The following segment will focus on factors which favour or hinder such engagement. In this regard, the analysis intents to shed light on general tendencies which can be observed with respect to the five different categories of transnational activities. Due to the large number of individual indicators, cumulative indices will be created for each of the subdimensions. These will then be used as dependent variables in linear or logistic multivariate regressions.

The index for transnational communication was built by summing up all three individual communication variables. These measured how often respondents communicate with friends and

152

The transnationalism of Turkish and Romanian migrants family members abroad via telephone or VoIP, via mail or email and via social networking sites. All three items used five-point scales with never (0) being the lowest category and every day (4) being the highest category. Consequently, the index can take values between zero and twelve. It is important to note that the items used to build the index referred exclusively to communication with friends and family in other countries and did not, for example, refer to professional cross- border communication with clients or employers. It should also be reiterated that the items were filter-dependent68 (for details see chapter 3.4). Consequently, respondents who indicated that they do not have any friends or family in other countries than the CoR, and therefore did not receive the communication questions, as well as those who did not provide any answer to the question for contacts abroad, were coded as missing and excluded from the analysis. Accordingly, the value zero is assigned to respondents who fulfil the basic criteria of having someone to talk to in another country but do nevertheless not engage in transnational communication. This approach is chosen since transnational communication serves as a dependent variable in the following analysis and can, due to the design of the survey, be explained to a great deal by the existence of friends and family abroad. As mentioned, this type of communication was only measured for respondents who reported social contacts in other countries. Consequently, the inclusion of those who did not report any such contacts in the base category (0) of the respective variables has in an additional regression (table not included) the effect that all network variables become highly significant, explain a great part of the variance and differences between the effects of the three variables on transnational communication are obscured.

The indicator for economic transnationalism is limited to sending remittances and having property in the CoO. This decision is based on the results of the multiple correspondence analysis discussed above (see Figure 6). The respective plot shows clear relations between the two mentioned indicators and a somewhat weaker relation between having property in a third country and receiving money from abroad. Hence, both pairs of indicators should be reviewed separately. However, at the same time, the descriptive analysis showed that the share of respondents who receive money or have property in third countries is very low. Consequently, both aspects are disregarded as dependent variables. The economic transnationalism index will, therefore, build on remitting behaviour and property in the CoO. For the remittances variable, all respondents who mentioned that they do not send any money abroad received the value zero. Furthermore, the

68 Only those respondents who indicated that they had private contacts abroad were asked how often they communicate with them.

153

The transnationalism of Turkish and Romanian migrants following three categories were used: sending remittances less than once a year (1), at least once a year (2) and at least once a month (3). The variable signalling the existence of property in the CoO was recoded so that the value two indicates property, while the value zero indicates its absence. This means that the same numeric value is assigned to the possession of real estate in the CoO and remitting money at least once a year. This reflects the assumption that property in the CoO will bring the respective respondent in direct contact with individuals or institutions in that country at least once a year, or more often, but most likely not every month. Direct contact in this context refers to cross-border interactions and exchanges which are closely related to the property itself. Such interactions could consist of conversations and arrangements concerning the maintenance of buildings, exchanges with tenants or family members using the property, or with state institutions, e.g., in the context of real estate taxation. Finally, the indicators for remittances and real estate ownership are combined in an additive index which can take values between zero and five.

The index variable measuring transnational mobility builds on two items which asked for the frequency of visits to the CoO and visits to third countries in the 24 months prior to the survey. In a first step, the raw variables were recoded so that their values could be organised into the following categories: no trips at all (0), one trip (1), two trips (2) and three or more trips (3). An additive index of both variables was created in a second step. The respective variable can, therefore, take values between zero and six, where zero means that the respondent did not take any trips abroad, and six indicates that he or she visited the CoO three times or more whilst also travelling at least three times to a third country.

Political transnationalism is measured by a dummy variable which indicates whether the respondents participated in the last CoR election. Respondents who were not eligible to vote or did not yet live in the CoR at the time of the last election were coded as missing and are not included in the analysis.

The index measuring transnational consumption is based on three dummy variables indicating whether respondents generally watched foreign-language TV content, purchased goods, and received any electronic messages from abroad during the 12 months preceding the survey. It should once more be stressed that, in contrast to how transnational communication was operationalised, the third variable used in the transnational consumption index indicates whether

154

The transnationalism of Turkish and Romanian migrants respondents received emails or social network messages (excluding spam) from abroad. While receiving such messages indicates exposure to border-crossing content, it does not automatically mean that the respondents engaged in active conversations. All three indicators of transnational consumption are combined into a compound variable which can take values between zero and three.

Table 9 presents two models which were computed for each type of transnational practice. The first model concentrates on various sociodemographic factors, discrimination experience, and group identifiers. Model 2 introduces additional controls for transnational background and networks, and for transnational human capital. This strategy will allow for the observation of the degree to which both dimensions influence transnational practices.

The independent variables used in the following regressions are described in Appendix 6. For an overview of all hypotheses see Table 1.

155

Table 9: Determinants of transnational practices (coefficients)

Model 1 Model 2 Transnational practices Transnational practices

Commu- Consump- Commu- Consump- Economic Mobility Political Economic Mobility Political nication tion nication tion

Gender (female = 1) 0.271 0.006 -0.093 -0.361 -0.185*** 0.365 0.088 0.008 -0.412* -0.163** Married/ registered civil union (ref.) Single, never been married 0.154 -0.271* 0.092 -0.225 0.208** 0.055 -0.287* 0.023 -0.281 0.145* Other 0.666* -0.132 -0.224 -0.660 0.196* 0.554 -0.184 -0.254 -0.664 0.150 Age at migration -0.048*** 0.022*** -0.002 0.025* -0.016*** -0.041*** 0.022*** 0.000 0.025 -0.012*** Duration of stay -0.072*** 0.003 0.000 0.027* -0.016*** -0.075*** 0.003 0.001 0.024 -0.016*** Education

Lower secondary or less (ref.) In-between lower and higher 0.751* 0.047 0.579** -0.088 0.350*** 0.707 0.014 0.553** -0.174 0.250* secondary education Higher secondary education 0.932*** -0.052 0.561*** 0.079 0.438*** 0.875*** -0.066 0.543*** 0.120 0.349*** Tertiary education 1.332*** -0.257* 1.308*** 0.580* 0.844*** 1.048*** -0.273 1.162*** 0.485 0.600*** Labour market participation

Working (ref.) Non-Working -0.016 -0.655*** -0.344* 0.343 -0.068 0.054 -0.651*** -0.311* 0.463 0.001 Retired -0.263 -0.705*** -0.293 0.077 -0.115 -0.031 -0.676*** -0.195 0.160 -0.019 Unemployed -0.068 -0.413** -0.265 0.206 -0.021 -0.068 -0.359* -0.182 0.269 0.021 Other 0.294 -0.516 -0.440 -0.156 0.019 0.231 -0.521 -0.569 -0.117 -0.038 Household financial sit. at age 14 0.125 -0.042 0.064 0.050 0.023 0.123 -0.040 0.044 0.022 -0.005 Household financial sit. now 0.302** 0.198*** 0.305*** 0.108 0.039 0.283* 0.171** 0.287*** 0.129 0.022 Property in the CoR 0.168 0.134 0.226* -0.080 0.131* 0.243 0.131 0.245* -0.056 0.127* Discrimination experience -0.004 0.154 0.041 -0.408* 0.077 -0.056 0.112 0.053 -0.350 0.078 Groups

TU in DK (ref.) TU in GE -0.291 -0.111 -0.253 0.341 -0.448*** -0.283 -0.074 -0.130 0.372 -0.353*** TU in IT 0.239 -0.136 0.114 -0.378 -0.376* 0.184 -0.134 0.058 -0.298 -0.372* RO in DK -0.406 -0.273 0.303 0.572 0.315** -0.717* -0.304 0.251 0.437 0.173 RO in GE -0.096 0.032 0.115 -0.530 -0.164 -0.381 0.040 0.159 -0.613 -0.213* RO in IT -0.120 0.043 0.097 0.595 -0.231* -0.288 0.061 0.160 0.586 -0.238*

Model 1 Model 2 Transnational practices Transnational practices

Commu- Consump- Commu- Consump- Economic Mobility Political Economic Mobility Political nication tion nication tion

Social contacts in other countries from CoO in CoO -0.055 -0.023 0.105 -0.031 0.002 from CoO in 3rd country 0.241 0.179* 0.070 -0.105 0.059 from 3rd country living abroad 0.325* 0.011 0.186* 0.288 0.197*** Partner lives in other country 0.150 0.629** 0.459 1.044** 0.041

Knowledge of a 3rd language 0.591** -0.027 0.255* 0.035 0.406***

(at least moderate) Knowledge of 3rd countries 0.135 0.081 0.271*** 0.107 0.096** Previous migration experience -0.164 0.038 0.163 -0.044 0.127

Constant 5.725*** 0.626 0.968* -3.592*** 1.345*** 5.165*** 0.517 0.281 -3.645*** 1.085***

Observations 1,102 1,347 1,341 1,191 1,295 1,027 1,260 1,258 1,115 1,220 Adjusted R-squared 0.156 0.089 0.160 0.358 0.177 0.096 0.200 0.413 Pseudo R-squared 0.0560 0.0631 *** p<0.001, ** p<0.01, * p<0.05. Political (transnationalism): logistic regressions. All other models: linear regressions. Additional controls: survey mode (all models). EUCROSS, 2016.

The transnationalism of Turkish and Romanian migrants

The next part of this chapter will discuss the results presented in Table 9 by successively focusing on various predictors. In order to facilitate the comparison between text and table, all dependent variables (i.e., the transnational practices) will be printed in italics. The possible influence of transnational human capital and transnational background and networks on transnational practices is explored at the end of this section. The structure of this section is chosen to enable a direct comparison of effects across transnational practices and to avoid redundancy.

Gender The first models in Table 9 show no gender effect for four out of five types of transnational activities. The only case in which there is a significant effect is transnational consumption. However, in the second set of regressions, i.e., if additional controls for transnational networks and transnational human capital are included, the dummy variable for female respondents also shows a significant negative relation with political transnationalism. Consequently, HT1.1 is supported by the data since men are more likely to participate in absentee voting. This corresponds to earlier findings, such as those of Guarnizo et al. (2003, 1235) and Dekker and Siegel (2013, 12). However, there is no support for the related second hypothesis (HT1.2) which predicted that women were in turn more likely to engage in the less formal activity of transnational communication. No significant gender effects are visible with regard to transnational mobility and economic transnationalism.

As mentioned, transnational consumption stands apart from the other four dimensions: while these activities are, on the one hand, not highly formalised, men – and, more precisely, single men – are, on the other hand, more likely to engage in these practices than women. Hence, this is the opposite of the expected effect. However, this deviation might also be related to the nature of the activities in question. In comparison to the other dimensions of transnationalism, transnational consumption is least related to migrants’ CoO, their interest in CoO affairs, and their family relations. In other words, of the five investigated practices, transnational consumption is the one where being a migrant could be assumed to be – in direct comparison – least relevant for the probability of engaging in it. Furthermore, it is remarkable that both models have, by far, the highest explanatory power for this type of transnational practice which also indicates that transnational consumption differs in its nature from the other four types of transnational practices.

158

The transnationalism of Turkish and Romanian migrants

Age at migration and duration of stay Four types of transnational activities show significant effects of the age at which respondents migrated to the CoR in Model 1 and three do so in Model 2. The positive impact of age at migration on economic transnationalism is the only of these significant relations which is present without a corresponding effect of the duration of stay. Hence, the data at hand show that engagement in economic transnationalism is the higher, the later in their life migrants settle in their CoR. This finding is in line with the result reported by Schans (2009, 1173) and contrasts those of Schunck (2014, 243) as well as Dekker and Siegel (2013, 12), who did not find corresponding effects. The descriptive analysis has shown that only a small part of all samples had previously lived in a third country (see Table 7), i.e., most of the respondents migrated from their country of birth. Consequently, this finding indicates that economic transnationalism is strengthened by the lifetime spent in the CoO and the resulting degree of socialisation in this environment. This interpretation is supported by the fact that previous migration experience, which would indicate that the respondents spent significant time elsewhere before settling in the CoR, is not significant. Furthermore, the significance level of age at migration remains the same in Model 2 as in Modell 1 which means that this effect works independently of transnational background and transnational human capital.

Concerning transnational consumption both time-related indicators (i.e., age at migration and duration of stay) viewed together would suggest that respondents who migrated at a younger age and have spent less time in the CoR are more likely to engage in this practice than those who migrated at an older age and a longer time ago. However, as discussed above (see chapter 3.4) it has to be taken into consideration that these two independent variables taken together are indicative of respondents’ age. Hence, three additional regressions of Model 1 and 2 were computed for transnational consumption using either only age instead of the two original predictors, or age in combination with age at migration or duration of stay (tables not included). All six of these computations show highly significant negative relations between age and transnational consumption while age at migration and duration of stay are not significant when included. In the interpretation of this finding it should be kept in mind that the included consumption practices are – to different degrees – either reliant on electronic communication media and the internet, or are at least substantially facilitated by them. It is, for example, easier to order something from abroad via the internet than by calling a company in another country.

159

The transnationalism of Turkish and Romanian migrants

Consequently, it could be assumed that the reported age effect actually points to a declining use of – or a decreasing familiarity with – such technologies with increasing age.

For transnational communication, six additional regressions corresponding to those mentioned above were computed to disentangle the combined effect of age at migration and duration of stay as well. This means that, again, age was used alone or in combination with either of the other two time-related variables in models otherwise identical to those presented in Table 9 (tables not included). The three alternative computations of Model 1 show the same age effects for transnational communication as described for transnational consumption. Hence, neither age at migration nor duration of stay are significant when combined with age while the latter indicator is negatively related to transnational communication and significant when combined with either of the other two predictors or used alone. Age is also negatively related to the dependent variable in all three alternative computations of Model 2. Hence, according to the analysis higher age is negatively related to transnational communication, which, in its intense form, is heavily reliant on the use of comparatively new technologies and internet services (such as social networking sites and VoIP software), too. This supports the mentioned assumption that the visible effects could indeed indicate a declining familiarity with communication technologies related to higher age. However, in the alternative regression of transnational communication which combines the duration of stay with age and additionally controls for transnational background and human capital (i.e., similar to Model 2), the duration of stay is still significantly and negatively associated with the dependent variable. Furthermore, in the regression of Model 2 which combines age and age at migration both indicators are significant but the coefficient of age is negative while the coefficient of age at migration is positive. In this context, it is relevant to recall that the duration of stay can be computed by subtracting age at migration from overall age. To summarise: The negative effect of duration of stay on transnational communication which is observed in Model 2 remains when age is controlled for, instead of age at migration. Furthermore, if the latter two variables are combined in the model – and duration of stay is not included – their effects also hint at an underlying negative duration of stay effect. Accordingly, it can be concluded that, at least in Model 2, the observed negative effect of duration of stay works

160

The transnationalism of Turkish and Romanian migrants independently from age.69 This indicates that the degree of transnational communication indeed declines the longer the respondents stay in the CoR.

The six additional regressions of political transnationalism (tables not included) point again in the direction of an age effect. If transnational background and human capital are not included in the regressions (i.e., similar to Model 1), age is significant in combination with either of the other two variables (which in turn are not significant) and if included alone. If, however, transnational background and human capital are controlled for (i.e., similar to Model 2), the only variable which is significant is age and only in a regression which does not include either age at migration or duration of stay at the same time. This means that if age is combined with either of the other two variables in a model which does control for transnational background and human capital, none of the three indicators (i.e., age, age at migration and duration of stay) is significant. Hence, age is more decisive than age at migration or duration of stay when it comes to political transnationalism. Furthermore, the results indicate that older immigrants are indeed more likely to participate in CoO elections than younger ones.

In sum, HT2 which postulated that the engagement in transnational practices would decline the longer respondents live in the CoR is neither supported regarding transnational mobility nor with respect to economic or political transnationalism. Regarding transnational consumption, there seems to be support for this hypothesis at first sight: The respective indicator is indeed negative and significant in both models presented in Table 9. However, the presence of a corresponding effect of age at migration has suggested and further analyses have demonstrated that these associations are more correctly interpreted as indications of an age effect.

Age is also a confounding factor in the case of transnational communication. Nevertheless, this is the transnational practice for which a negative effect of the time in the CoR is consistent across all analyses. Hence, the hypothesis postulating a negative relation between the time spent in the CoR and transnational engagement (HT2) is supported with regard to transnational communication.

69 Additional regressions of Model 2 which use either age at migration or duration of stay alone show no significant effect of the former but a significant negative relation of the latter with transnational communication (tables not included).

161

The transnationalism of Turkish and Romanian migrants

Education The relation between formal education and transnational activities appears to be rather heterogeneous. Guarnizo and co-authors (2003, 1216) point out that empirical research has generally shown a positive relation between higher education and increased political interest or a tendency to be politically active. Model 1 lends support to hypothesis HT3 with regard to political transnationalism in so far as tertiary education has a significant positive impact on the dependent variable. However, this also means, first, that only respondents who hold educational titles in the highest category differ from those who are in the lowest category, while the two intermediary educational levels are not significant. Second, Model 2 shows that even the tertiary education indicator is not significant anymore once transnational human capital and background are controlled for. Hence, the influence of education on political transnationalism is mediated by indicators of transnational human capital as well as transnational background and networks included in Model 2. This is mirrored by a similar, yet negative, effect of tertiary education on economic transnationalism in the first model which means that those respondents were significantly less likely to engage in border-crossing economic activities than the individuals in the reference group. However, just as in the former case this association between education and the dependent variable is not significant anymore if transnational human capital and background are controlled for (M2). Hypothesis HT3, which foresaw a positive relation between the engagement in transnational activities and formal education, is broadly supported regarding transnational communication, mobility, and consumption. For these dependent variables, all higher levels of education are significant in both models. It is noteworthy that the practices which are more common among the higher educated (i.e., transnational communication, mobility, and consumption) are all by default comparatively time-consuming and, to some extent, also have the character of spare time and leisure activities. Most notably this is the case with respect to such activities which either by definition entail social interactions (i.e., transnational communication) or which are very likely to lead to them (i.e., transnational mobility).

In a nutshell, the data show that respondents with higher levels of formal education are more likely to communicate, travel, and consume transnationally, while no consistent effects of education are visible with regard to economic and political transnationalism. This means that the relation between education and transnationalism is not uniform across different types of

162

The transnationalism of Turkish and Romanian migrants activities. Therefore, conclusions drawn from earlier research are only partially supported by the current analysis.

Labour market participation Being in full or part-time employment is used as a reference category for labour market participation in Table 9. This means that significant negative coefficients of the other statuses indicate that the respective respondents are less active transnationally than those who are working in formal employment. With respect to economic transnationalism Models 1 and 2 confirm HT4, given that besides absence from the labour market caused by chronic illnesses and disabilities (i.e., the category other) all statuses show a significant negative relationship to economic transnationalism.

For transnational mobility, both models show negative coefficients for all statuses in comparison to the working category with the non-working status being significant. This means that the respondents who are, for example, in education or care for children, and, therefore, do not participate in the labour market, constitute the only group which is significantly less likely to engage in transnational mobility than respondents who are in employment. While indicators of absence from the labour market showed comparable negative associations in Schunck’s model, none of the variables was significant in his analysis (Schunck 2014, 224). It can be assumed that the inclusion of visits to third countries in the present measurement contributed to these differences. More importantly, these results also mean that being employed and the time constraints possibly resulting from it do not have a negative impact on this type of cross-border activity.

Regarding the other dimensions, the results presented in Table 9 are mostly in line with those previously reported by other authors who did not find any significant effects of labour market integration on transnational communication (Schans 2009, 1176) or political transnationalism (Dekker and Siegel 2013, 11).

Household financial situation No significant relations are visible between political transnationalism or transnational consumption, and the subjective assessment of the respondents’ current household financial situation. This corresponds to the absence of significant effects of labour market integration on these two transnational practices.

163

The transnationalism of Turkish and Romanian migrants

With respect to the three other types of practices, a positive effect can be observed, i.e., the more positively respondents describe their household financial situation, the more likely they are to engage in these transnational activities. This effect has a higher significance for the two forms of transnationalism which directly depend on financial means, namely economic transnationalism and transnational mobility.

As described above, in addition to the household financial situation, labour market participation is also very relevant with respect to economic transnationalism but to a lesser degree when it comes to transnational mobility. A possible interpretation is that the predictability of one’s economic situation is important with respect to economic transnationalism, besides the current financial situation respondents find themselves in. The assumption underlying this argument is that individuals who are in employment usually expect that this employment, and the associated income, continues in the future. The importance of assumed future economic stability also appears reasonable since economic transnationalism includes caring for property and, as the descriptive analysis above showed, financial commitments made primarily to family members. Hence, these activities are by nature long-term oriented. Furthermore, a clear negative relation between being single and economic transnationalism is visible in both models. Seen in combination with the importance of the household financial situation, this could mean that individuals who can only rely on their personal income are less likely to engage in economic transnationalism than those who are in a committed relationship and, therefore, are more likely to be financially supported in their efforts by their partners, be it directly or indirectly. Mobility, on the contrary, does usually not entail a long-term commitment of a financial nature but is much more dependent on the availability of resources at a particular point in time. Thus, the household financial situation at a given point in time is of stronger relevance in this case than the question of its prospective development, which could explain why significant effects of labour market integration are largely absent in the case of transnational mobility. Hence, the presence and combination of different effects regarding those two dependent variables could be related to the question of whether or not these activities are likely to constitute a long time commitment. It should, however, be stressed that this interpretation cannot be tested with the data at hand. The negative relation between the single status and economic transnationalism could, for example, also mean that married individuals are more committed to conservative family values and, therefore, tend to financially stronger support family members abroad. The nature of the positive

164

The transnationalism of Turkish and Romanian migrants relation between transnational communication and household income is not straightforward, as transnational communication does not require large monetary investments.

In conclusion, the presented results support HT5 given that respondents are more likely to engage in economic transnationalism the better they assess their current household financial situation to be. Furthermore, such a positive assessment is also positively related to transnational communication and mobility, whereas no significant effects could be found regarding political transnationalism and transnational consumption. Hence, in this respect, the discussed findings are in line with those of other authors who found considerable variance in the significance of income-related indicators across different types of transnational activities.

Property in the CoR Significant effects of real estate ownership in the CoR can be observed with respect to transnational mobility and consumption. Regarding transnational mobility, this is in agreement with the reported significance of the household financial situation on this practice, as both predictors indicate some form of financial stability. Accordingly, the other two types of transnationalism in which the current household financial situation is significant (i.e., transnational communication and economic transnationalism) also show positive, but not significant, effects of property ownership. Hence, hypothesis six (HT6) is supported with regard to transnational mobility. However, no significant effect can be found regarding economic transnationalism. These findings, first, stress again that factors influencing transnational engagement differ with respect to the various types of activities. Second, the results contradict the mentioned claims by Huntington (2004, 272), as no signs of a negative relation between economic transnationalism and the ability or willingness to invest in CoR property are visible in the data.

Discrimination experience The fact that the indicator of discrimination experience does not show significant effects in all but one of the regressions means that this data does not support the reactive transnationalism hypothesis brought forward by Itzigsohn and Giorguli Saucedo (2002). This hypothesis suggested that perceived discrimination would significantly enhance migrants’ engagement in transnational practices (see chapter 2.3). Indeed, the only type of transnational practice for which a statistically significant relation with discrimination is visible is political transnationalism, as long as neither

165

The transnationalism of Turkish and Romanian migrants transnational human capital, nor transnational background and networks are controlled for (i.e., Model 1 in Table 9). However, contrary to the reactive transnationalism hypothesis, the effect is negative instead of positive. This means that perceived discrimination is not associated with an increase but decrease in the participation in CoO elections. In their text, Itzigsohn and Giorguli Saucedo (2002, 783) additionally interpret the positive relation between higher education and transnationalism as a sign of reactive transnationalism, arguing that better-educated migrants are more likely to encounter discrimination and, consequently, more likely to engage in cross-border activities. The authors state that this corresponds to a bivariate relation between discrimination and education in their dataset. In the present case, however, an additional computation (table not included) showed that discrimination is not positively related to higher education. Hence, the reported positive effect of higher education should not be seen as an indicator of reactive transnationalism either.

Groups While the descriptive analysis revealed substantial differences between migrants of both nationalities in Denmark, Germany, and Italy, remarkably few effects of group-dummies on transnational activities are visible in either of the models presented in Table 9. In fact, transnational consumption, i.e., the type of transnationalism which is the least related to migration, constitutes the only dimension for which both models show significant differences between migrant groups. Only with respect to transnational communication an additional effect of the group indicator of Romanians in Denmark is visible in Model 2. It indicates that these respondents engage less in this practice than Turks in the same country if social networks and language knowledge are controlled for. Nevertheless, on a general level, these results show that other predictors have a more significant impact on transnational practices than the combination of migrants’ nationality and CoR. Hence, this analysis indicates that engagement in four dimensions of transnational activities is not strongly influenced by context factors at a national level, i.e., most importantly by the sum of CoR policies directed towards all migrants or specifically concerning Turkish or Romanian migrants. This does not, however, exclude the presence of contextual factors in general. On the one hand, it has to be taken into account that the indicator used here is by its very nature a global one (i.e., the grouping by countries controls for the sum of context effects), and does not single out specific policy areas and compare their impact across

166

The transnationalism of Turkish and Romanian migrants countries. On the other hand, there could also be important influencing factors on a subnational level which are also not captured by the models in Table 9.

Additionally, it is worth considering the coefficients of the group-dummies in the regression of transnational consumption in Model 2 more closely: In Denmark, Turkish and Romanian migrants do not differ significantly from each other. However, respondents in all four of the other groups are significantly less likely to engage in transnational consumption. Furthermore, the coefficients of Turks in Germany and Italy are quite similar to each other, which is also true for those of Romanians in these countries. Taken together, this suggests that the observed difference for transnational consumption is less influenced by the respondents’ nationality, than by their CoR. Two additional sets of regressions resembling Model 2 were computed for all types of transnational activities to review this effect (tables not included). Instead of the above-used combination of CoR and CoO, the first set of regressions included a variable differentiating only between respondents’ nationality, while the second set differentiated only between the countries in which they resided. While the first set did not show effects of nationality for any of the transnational practices, the dummies distinguishing migrants in Germany and Italy from those in Denmark showed significant negative coefficients in the regression of transnational consumption but not for any of the other practices. Hence, the above-made interpretation is supported by these additional results. Differences between groups are only visible in regard to a single type of transnational activity. Furthermore, they do not appear between the two nationalities but are connected to contextual factors related to the CoR. However, the observed differences between respondents in Denmark and the other two countries are most likely not caused by structural differences related to the specific situation of migrants. Instead, the effect is likely fostered by the fact that people in Denmark have more easy access to foreign-language TV content than those in the other two countries (see descriptive analysis above). Additionally, Denmark is a comparatively small and, at the same time, highly globalised market which might also foster cross-border shopping on an individual level. In this sense, the different aspects included in the consumption dimension could also enhance each other: If someone buys something online in another country, he or she is also more likely to receive electronic messages from abroad.

As indicated, these findings do, of course, not mean that no relevant differences regarding the degree of transnational engagement exist between the six subsamples. The descriptive analysis showed that this is indeed the case. However, the results of the discussed regressions imply that

167

The transnationalism of Turkish and Romanian migrants these differences cannot chiefly be explained by factors which pertain to either the CoO or the CoR of respondents but have to be attributed to other predictors, such as those of a socio- economic nature.

Transnational background and networks and transnational human capital Transnational background and networks, and transnational human capital constitute the other two key dimensions of transnationalism besides transnational practices (see chapters 2.3 and 3.4). It has been argued above that aspects of all three dimensions are likely to influence each other. Consequently, the respective predictors shall be discussed jointly in the following.

Besides socio-economic aspects, transnational communication is positively related to third country contacts abroad and a certain level of additional language knowledge. At the same time, neither of the variables that measure the size of networks of individuals from the CoO in different locations is significant. This indicates that the degree to which migrants engage in transnational communication is not only influenced by the extension of transnational networks in terms of individuals connected by them but also by their composition, or more precisely their diversity. Hence, those respondents whose networks are more diverse differ systematically in their transnational communication behaviour from those who have more homogeneous networks in regard to the origin of the friends and family members forming part of them.

Having a life partner in another country – which constitutes an additional aspect of transnational background and networks – is positively associated with economic and political transnationalism. It can be assumed that life partners, in most cases, live in the respondents’ CoO. Hence, the respective effects are indicative of a continuing close personal connection with the CoO. The model also shows a significant positive effect between economic transnationalism and the size of networks consisting of CoO individuals living neither in the CoO nor in the CoR. This indicates that economic transnationalism is more likely to occur if the respondents are part of larger migration networks. On the one hand, it is possible that the transnational engagement of some members, e.g., in the form of remittances, within such networks might create social pressure on others to become active in this regard, too. On the other hand, it is also conceivable that economic transnationalism within such networks might provide individuals with social prestige if they show the ability to remit or invest money across borders, while other members of the network are not able to do so, or can only afford such activities in a more restrictive manner.

168

The transnationalism of Turkish and Romanian migrants

However, the results show no support for HT7 which assumed that the number of friends and family who are of CoO origin and also live in the CoO would be positively related to economic transnationalism. It follows from the above that instead of the quantity of contacts in the CoO it is the quality of this relation, i.e., the degree of personal closeness between the migrant and the individual in the CoO, which seems to impact economic transnationalism.

Furthermore, Model 2 shows that having a higher number of acquaintances with a third country background who live abroad is positively related to transnational mobility. Additionally, this practice is also related to transnational human capital, as the indicators of profound knowledge of other countries besides the CoO and CoR, and at least decent knowledge of a third language, are both significant. Generally, it can be assumed that these relations are mutually reinforcing: A larger network abroad provides possibilities to visit other countries which might, in turn, lead to new contacts. In the same way, the knowledge of third countries, as well as languages, might have a positive effect on mobility, which could lead to higher language capacity and broader understanding of foreign countries.

Finally, aspects of transnational human capital show highly significant relations with transnational consumption. This is not surprising since, for example, it seems self-evident that knowledge of a third language would – if anything – enhance the probability of consuming foreign TV content and cross-border shopping. In the same way, the knowledge of a third country might both enhance the interest in this country – which might lead to corresponding media consumption – and the familiarity with country-specific products which, again, could increase the probability of cross-border shopping.

On a general level, a comparison of the explained variance of both sets of models shown in Table 9 confirms that transnational background and networks, as well as transnational human capital partially explain the degree to which respondents engage in transnational activities. While different aspects of both dimensions are significantly related to transnational activities, the adjusted R² or the pseudo-R² is for all dimensions somewhat higher in the second model. However, while the increases in the explained variance are apparent, they are not extreme in their extent. The largest effects can be observed for transnational consumption, mobility, and communication where they range between 5.5 and 2.1 percentage points.

169

The transnationalism of Turkish and Romanian migrants

5.6 Summary This chapter started out by confirming the relation between indicators of the different dimensions and subdimensions of transnationalism through multiple correspondence analysis. This empirical justification of the used multidimensional operationalisation of transnationalism does not, of course, exclude the possibility of systematising the phenomenon in other ways. In fact, the presented operationalisation is based on several such approaches. First and foremost to be mentioned in this regard are the approach by Kuhn (2011) and the categorisation of transnational activities used by collaborators of the CIEP project (e.g., Portes, Guarnizo, and Landolt 1999). The purpose of the exercise was, however, to provide empirical evidence for the theoretically assumed relations between the indicators within each dimension.

During the following examination of these dimensions, it became apparent that the majority of respondents are part of transnational social networks, even though the composition of these networks and their mean size differs between the samples. Generally, the personal networks of the interviewed migrants show some striking similarities: First, transnational networks are spanning a space that, for the majority of respondents in all but one sample, clearly goes beyond a simple CoR-CoO dichotomy. Even amongst Turks in Germany, who constitute the one case in which this does not hold true, nearly every second respondent has friends or family in a third country. Second, the networks of migrants from both nationalities are largely centred on Western Europe. Finally and unsurprisingly, there is a relation between the size of the respective migrant community in a third country and the probability of migrants to have connections to it. Hence, more than 60 percent of Turkish migrants in Denmark and Italy have friends or family in Germany, which hosts the biggest Turkish community outside Turkey. In principle the same applies to Romanian migrants, however, there it is not one country but a triad of Italy, Spain, and Germany which is most likely to host respondents’ friends and family besides the CoO and the CoR. This triad, in turn, corresponds to the higher degree of geographic dispersion of the Romanian community abroad, which features these three countries as most important destinations (see chapter 4). The analysis of respondents’ geographic patterns of physical mobility, just as in the case of transnational networks, provides evidence that it is crucial to go beyond the CoR-CoO dichotomy present in migration research in order to understand the true extent of migrant transnationalism. While the CoO constitutes the main destination, it is by no means the only country visited by migrants. The list of often visited destinations is, furthermore, not limited to

170

The transnationalism of Turkish and Romanian migrants countries which respondents identified as the main settlement regions of their friends and family abroad. In contrast to the social networks, mobility patterns include a wider array of European countries; however, they are also more Eurocentric, as no countries outside the continent were mentioned as destinations by at least five percent of any of the samples. The most important mobility reasons align with earlier results on sedentary Germans presented by Mau (2007), given that vacations and visits to family and friends were stated most often in both cases.

It has been shown that transnational communication is clearly part of the everyday life of most of the respondents. The results revealed that the interviewed migrants engage in it, on average, at least once a week. For doing so, traditional telephone conversations or VoIP calls are the preferred mode. This means that respondents mostly rely on verbal forms of communication, while non-verbal forms are less intensively used. In contrast to transnational communication and mobility, fewer respondents engage in economic transnationalism and transnational consumption. It has been shown that the share of migrants in the analysed samples who remit money is lower than most of the figures reported in other studies. This is especially true in comparison to the results from the United States. Besides methodological differences which might partially account for the discrepancies, these figures, more importantly, underline that the degree to which migrants engage in monetary transfers vary, not only between different settlement countries but also between groups from different CoOs within the same CoR. The same holds true with regards to the existence of property in the CoO. These differences do, however, not mean that these transnational practices are determined by group membership (see below). Generally speaking, the share of respondents who hold property in the CoO is considerable in all samples, ranging from one in four Romanian migrants in Denmark to nearly every second Romanian participant in Italy. Migrants are much less frequently recipients of money send to them from abroad or owners of property in a third country. Political transnationalism, measured in terms of absentee voting, is with slightly more than 14 percent on a comparable level to the results reported by Waldinger (2008) for immigrants in the United States.

In general, both the descriptive and the regression analyses underlined that it is beneficial to differentiate between various dimensions of transnationalism. On the one hand, it has been shown that each sample differs in the degree of its members’ transnational background and networks, their transnational human capital, and their engagement in transnational practices within the distinguished categories. On the other hand, it has been demonstrated that the same predictors

171

The transnationalism of Turkish and Romanian migrants varied in their relation to transnational activities of different types. This means that the various transnational practices are influenced by different sets of independent variables. In fact, none of the predictors in the second model is significantly related to more than three of the categories of transnational practices and many show only effects in one or two of them. It is doubtful that these nuances would have been visible if a less comprehensive measurement would have been used.

The presented findings do not lend support to the reactive transnationalism hypothesis formulated by Itzigsohn and Giorguli Saucedo (2002) since none of the dependent variables in Table 9 shows a significant positive relation with discrimination experience or is influenced by a negative assessment of the respondents’ household financial situation.

The descriptive analysis showed various differences regarding the level of transnationalism between the six samples; however, in the regressions of four out of five types of transnational activities, significant effects of the group indicators were almost completely absent. This means that the degree to which respondents engage in these practices is not determined by factors that can be generally attributed to the respondents’ CoO and CoR. Hence, while CoO and CoR were – for practical reasons – relevant categories of the sampling procedure, they do not constitute significant analytical categories as determinants of transnationalism. Instead, the individual level of engagement in transnational activities is related to diverse sets of factors which are in part migration specific (such as duration of settlement) but in part also of a more general socio- economic nature (e.g., current household financial situation). Indeed, economic engagement and cross-border mobility can, for example, be described as resource dependent forms of transnationalism in the sense of Itzigsohn and Giorguli Saucedo (2002).

The regression analyses showed the expected link between transnational practices and transnational background and networks as well as transnational human capital. Interestingly, through the inclusion of these aspects, it became, for instance, apparent that engagement in transnational communication is not only related to the existence and scope of transnational networks. In fact, the diversity of such connections is also highly relevant as indicated by the significant positive association between the existence of friends and family of third country origin, and the frequency in which respondents engage in cross-border communication. Furthermore, some effects reported in earlier research were only visible with respect to transnational practices when the two other dimensions (i.e., transnational human capital, and

172

The transnationalism of Turkish and Romanian migrants transnational background and networks) were controlled for. For example, this applies to hypothesis HT1.1 which stated that men were more likely to engage in political transnationalism than women. On a general level, Table 9 shows that the explanatory power of the presented model increases for all types of transnational practices when transnational background and transnational human capital are controlled for. This presents an additional argument for the systematic inclusion of these aspects in further analyses of transnational activities.

173

6 The relationship between transnationalism and integration

This chapter investigates the relationship between transnationalism and integration. As discussed in chapter 2, it does so by concentrating on social and identificational integration. With regard to the second, two distinct aspects – identification with the town and country that the respondents reside in – will be discussed separately. Each of the following sections will begin with a descriptive analysis of the respective integration indicators. Subsequently, a number of regressions will be presented which use these aspects as dependent variables and investigate how they are influenced by the various aspects of transnationalism discussed in chapter 5. At the end of the chapter, a comparison of findings regarding all three aspects of integration and a summary will be provided.

With the goal of drawing a differentiated picture of the interrelations between integration and transnationalism, the regressions will use the individual indicators of transnationalism instead of the indices discussed in the previous chapter. However, the effects of the indices in the overall models for each of the dependent variables will be briefly reviewed as part of the summary section at the end of this chapter. This will allow answering the question of whether their cumulative effects differ from those of the individual indicators.

6.1 Social integration Figures 17 to 19 show the distribution of social contacts amongst migrants of both nationalities in all three countries under observation. The first of these figures (17) presents the data of the variable used to operationalise social integration, that is the number of family members and friends of CoR origin respondents have. The second figure (18) presents comparable data regarding individuals of CoO origin, while the third (Figure 19) refers to friends and family who originated in a third country.

Regarding friends and family members of CoR origin, Figure 17 reveals that most respondents show some degree of social integration, since overall 82 percent of the Turkish and 77 percent of the Romanian migrants have such individuals in their personal networks. Interestingly, the total percentage of Turkish respondents who have such contacts is in all three samples consistently higher than amongst their Romanian counterparts. In other words: Respondents who do not have any close social contact with CoR natives are a minority in all three countries and amongst both

174

The relationship between transnationalism and integration nationalities. However, they constitute a somewhat bigger share of the Romanian than of the Turkish samples. It is also remarkable that, for both nationalities, the relative sizes of the subsamples without any close CoR contacts are practically identical in all three countries. The respective differences vary only by 1.2 percentage points in the case of both nationalities.70

Figure 17: Family, in-laws, and friends from the CoR living in the CoR (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 27 54

TU in DE 20 62 A lot TU in IT 47 36 A few RO in DK 26 51

RO in DE 32 45

RO in IT 34 43

Figure 18: Family, in-laws, and friends from the CoO living in the CoR (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 82 13

TU in DE 80 14 A lot TU in IT 54 35 A few RO in DK 50 39

RO in DE 56 29

RO in IT 56 35

Figure 19: Family, in-laws, and friends from a third country living in the CoR (percent)

0 10 20 30 40 50 60 70 80 90 100

TU in DK 11 46

TU in DE 8 32 A lot TU in IT 25 45 A few RO in DK 23 47

RO in DE 14 38

RO in IT 11 32

Figures 17-19: EUCROSS, 2016. N = 1,502.

70 More specifically, between 17.2 and 18.4 percent of the Turkish and between 21.6 and 22.8 percent of the Romanian respondents did not report any close social contact with people of CoR origin.

175

The relationship between transnationalism and integration

After reviewing evidence for the general existence of such social relations, it is also of interest to evaluate their density. When doing so, a certain shift in the above-mentioned pattern appears. This is most obvious in Germany where approximately one in three Romanian respondents has a lot of CoR friends and family members, in comparison to only one in five Turkish migrants. In Denmark, this group has roughly the same relative size in both samples. Only in Italy, the relative share of the Turkish sample that attested to having a comparatively high number of CoR friends and family members is 13 percentage points higher than that of the Romanian sample. This is, at the same time, also the highest difference between both nationalities. Consequently, a somewhat stronger degree of social integration is visible for a larger number of Romanian than Turkish respondents in Germany, while the contrary holds true in Italy, and few differences can be observed in this respect in Denmark.

After reviewing both nationalities in comparison, finally, differences within the national groups across CoRs should be mentioned. Whereas the overall percentage of respondents with friends and family of CoR origin is nearly identical within both nationalities, clear differences are visible in the amount of such social connections. As mentioned, only 20 percent of Turkish respondents in Germany reported the existence of a high number of close CoR contacts. Amongst their co- nationals in Denmark, this number was seven percentage points higher, while in Italy nearly half of all respondents reported a lot of CoR contacts, giving them the highest percentage of all six samples. Consequently, there is a difference of 27 percentage points between Turks in Germany and in Italy. While the Romanian samples differ regarding this aspect as well, the differences are less pronounced, with percentages ranging between 26 percent in Denmark and 34 percent in Italy. This means that for both nationalities, the respondents in Italy have most contacts within the CoR population.

Figure 18 shows that it is for all samples somewhat more common to have family members or friends of CoO rather than of CoR origin in the country they reside in. Differences in the total values are small but notable. Amongst Turkish respondents, it is slightly more often the case in Denmark and Germany than in Italy. For Romanians, it is most common in Italy and least common in Germany.

However, more pronounced are again the differences in the number of these social contacts. The values of the Turkish samples in the two Northern European countries are nearly identical in this

176

The relationship between transnationalism and integration respect, with 80 percent or more of the samples having a lot of friends and family from the CoO in their CoRs. These are also the two countries to which a significant proportion of the Turkish samples migrated for family or love related reasons (see Table 4), i.e., to be with family members or partners. With respect to Romanian migrants, indications of a relation between the number of CoO contacts in the CoR and family-related migration are much weaker. Family motives for migration were most important amongst Romanians in Germany. However, the share of the sample with a lot of close CoO contacts in this country does not differ compared to Italy, where love or family-related migration played a far less important role. Still, the sample in Denmark, in which family motives were least important in respect to the decision to migrate, also has the lowest number of Romanian respondents with a lot of acquaintances from their CoO. The same pattern is visible in the case of the Turkish sample in Italy. Aside from Turks in Denmark and Germany, the relative portion of the samples which reported a lot of co-national friends and family members in the CoR ranged between 50 and 56 percent of the respondents.

There is no direct relation between the total size of the migrant communities in each of the surveyed countries (see chapter 4) and the number of CoO contacts respondents have in the CoR. For example, the difference in the absolute numbers of Turkish residents in Denmark and Germany and their relative share in the population of both countries is considerable (Appendix 2). However, this is not reflected in the probability of Turkish respondents to have Turkish friends or family in the CoR in this survey. In fact, Turks in Denmark are slightly more likely to have Turkish friends and family in their CoR than their co-nationals in Germany, despite the fact that many more Turks live in the latter rather than in the former country. However, the lack of such a direct link is not surprising. While it might be true that the chances of meeting a random person of Turkish or Romanian nationality could be higher in a country where the corresponding group constitutes a bigger part of the total population, this does not mean that the same can be said for the presence of friends and family. In fact, it has been shown by previous research that co-ethnic networks – consisting of family members, friends or former neighbours – can help migrants to establish themselves in new countries (see, e.g., Goldring 1999, 173; Müller-Mahn 2000). The existence of, and reliance on, such networks also has the consequence that the individual would have a certain number of CoO contacts in the CoR. However, there is no reason to assume that such networks are more likely to be present in countries which host a particularly huge community of migrants from Turkey or Romania. In fact, the contrary might be true, as

177

The relationship between transnationalism and integration migrants could be more likely to rely on personal networks in countries in which no considerable ethnic networks exist.

The comparison of Figures 17 and 18 shows that more respondents have close CoO contacts in their CoR than friends and family members who are of CoR origin. However, these differences are modest, ranging between 7 and 14 percentage points in the different samples. However, if the comparison concentrates on the number of respondents in each sample who reported having a lot of social contacts the differences grow considerably. The only exception are Turks in Italy amongst whom the percentage is only seven percentage points higher in Figure 18 than in Figure 17. The share of respondents with many CoO contacts is between 22 and 24 percentage points higher in comparison to those having many CoR contacts in all three Romanian samples. However, in the remaining two Turkish samples these figures are considerably higher. The difference between the share of the sample reporting a high number of CoO and CoR nationals is at 55 percentage points in favour of CoO nationals in Denmark and at 60 percentage points amongst Turkish respondents in Germany. Hence, it is far more common for Turkish migrants in those countries to have a lot of friends and family members who share their origin, than to have a high number of acquaintances who are natives of their CoR.

It is worth mentioning that the number of respondents who said that they do not have any friends and family members back in the CoO is – in all samples – higher than the number of respondents who reported no such contacts in the CoR (compare Figure 7). Besides this general observation, the number of respondents who reported having a lot of CoO friends and family members in their CoR is for all samples – except for Romanians in Denmark – higher than the number of respondents who have a lot of co-national friends and family in their CoO.

Finally, Figure 19 shows that, with the exceptions of Turks in Germany and Romanians in Italy, more than half of the respondents in all samples have friends or family with a migration background different to their own. The number of respondents with such contacts is highest amongst Turkish migrants in Italy and Romanians in Denmark, both on a general level and with respect to those who have a lot of these individuals in their networks. These samples also have the highest mean level of education compared to the samples of the same nationalities in the other two countries. At the same time, those samples in which such contacts were least common are those with the lowest mean level of education. A regression (table not included) in which the

178

The relationship between transnationalism and integration indicator for family and friends from a third country living in the CoR was used as dependent variable and the level of education as independent variable, confirmed that both these aspects are positively related. This is not surprising, as higher education increases the chance of learning another language, attaining a higher labour market position and coming into contact with other cultures in private or professional contexts.

Even though Figures 17 to 19 were presented, to a certain degree, in a contrastive fashion it is important to underline that the three mentioned categories cannot be equated with distinctive social networks. On the contrary, it is possible that individuals of any of the three groups which were differentiated for measurement purposes could actually be members of the same social circles. This means the differentiation used here for analytical reasons does not necessarily correspond to empirically distinct groups.

Table 10 shows the relationship between social integration (i.e., the number of CoR friends and family members) and transnationalism. It consists of a total of eleven models. The first model (M1) only uses the dependent variable and group indicators as independent variables, while Model 2 (M2) introduces socio-economic and migration-related variables. Models three to nine (M3-M9) introduce the indicators of the various dimensions of transnationalism. Finally, Model 10 (M10) includes the indicator of identification with the CoO and Model 11 (M11) is a full model with all previously introduced variables.

The independent variables used in the following regressions are described in Appendix 6. For an overview of all hypotheses see Table 1.

179

Table 10: Ordinal logistic regression of social integration (coefficients) M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Groups TU in DK (ref.) TU in DE -0.191 0.054 0.103 0.056 -0.022 0.059 0.103 0.039 0.110 0.059 0.159 TU in IT -0.748* -0.869* -0.773* -0.846* -0.882* -0.758* -0.920** -0.864* -0.834* -0.885* -0.673 RO in DK -0.113 -0.230 -0.166 -0.305 -0.219 -0.216 -0.237 -0.226 -0.311 -0.232 -0.175 RO in DE 0.024 -0.264 -0.247 -0.339 -0.333 -0.234 -0.266 -0.242 -0.289 -0.266 -0.303 RO in IT 0.108 0.088 0.127 0.060 0.056 0.141 0.125 0.091 0.085 0.075 0.180 Gender (female = 1) 0.120 0.157 0.129 0.138 0.152 0.121 0.112 0.155 0.115 0.203 Married/ registered civil union (ref.) Single, never been married -0.065 0.005 -0.066 -0.063 -0.057 -0.067 -0.058 -0.079 -0.051 0.002 Other -0.331 -0.257 -0.304 -0.348 -0.316 -0.331 -0.324 -0.489* -0.316 -0.364 Age at migration 0.001 -0.002 0.004 0.001 0.004 -0.001 0.000 0.004 0.002 0.007 Duration of stay 0.013 0.014 0.013 0.015 0.016* 0.012 0.011 0.015 0.013 0.017 Education Lower secondary or less (ref.) In-between low. and higher second. 0.247 0.373 0.230 0.271 0.226 0.226 0.274 0.315 0.237 0.396 Higher secondary 0.045 0.093 -0.003 0.048 0.059 -0.023 0.047 0.066 0.039 -0.000 Tertiary 0.300 0.403* 0.193 0.301 0.307 0.229 0.318 0.229 0.308 0.205 Received highest education in CoR 0.002 -0.091 0.003 -0.001 0.009 -0.054 0.004 -0.043 0.028 -0.079 Labour market participation Working (ref.) Non-Working -0.012 0.060 0.004 -0.012 -0.023 0.037 -0.005 -0.007 -0.013 0.103 Retired 0.084 0.135 0.089 0.076 0.003 0.125 0.105 0.135 0.074 0.095 Unemployed -0.275 -0.220 -0.253 -0.293 -0.304 -0.246 -0.275 -0.256 -0.267 -0.155 Other 0.354 0.332 0.334 0.333 0.276 0.366 0.368 0.293 0.362 0.149 Household financial sit. at age 14 -0.010 -0.023 -0.006 -0.022 -0.006 -0.021 -0.016 -0.035 -0.012 -0.042 Household financial sit. now 0.191** 0.196** 0.195** 0.179* 0.210** 0.171* 0.191** 0.214** 0.188* 0.209** Property in the CoR 0.015 0.028 0.050 0.013 0.035 -0.028 0.023 -0.033 0.014 0.015 Discrimination experience -0.309** -0.288* -0.313** -0.317** -0.290* -0.321** -0.289* -0.272* -0.311** -0.256* Knowledge of other CoR regions 0.187** 0.148* 0.187** 0.166* 0.199** 0.175** 0.193** 0.192** 0.187** 0.171* Knowledge of the CoR language No or poor knowledge (ref.) Intermediate knowledge 0.248 0.246 0.227 0.271 0.245 0.266 0.207 0.223 0.254 0.197 Good or very good knowledge 0.799*** 0.778*** 0.767*** 0.774*** 0.773*** 0.814*** 0.760*** 0.755** 0.802*** 0.681** CoR partner 0.876*** 0.864*** 0.807*** 0.873*** 0.910*** 0.886*** 0.847*** 0.898*** 0.887*** 0.872*** Social contacts in the CoR from CoO 0.179* 0.147 0.178* 0.185* 0.197* 0.178* 0.176 0.186* 0.177 0.192 from 3rd country 0.801*** 0.765*** 0.809*** 0.800*** 0.813*** 0.776*** 0.793*** 0.799*** 0.807*** 0.768***

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Social contacts in other countries from CoO in CoO 0.054 0.057 from CoO in 3rd country 0.221* 0.253* from 3rd country living abroad 0.047 0.023 Partner lives in other country -0.297 -0.219 Knowledge of a 3rd language 0.223 0.171 (at least moderate) Knowledge of 3rd countries 0.050 -0.007 Previous migration experience -0.246 -0.203 Frequency of transnat. communication via phone and VoIP 0.082 0.096 via mail and email 0.111* 0.100 via social netw. sites -0.030 -0.087 Frequency of remittances -0.003 -0.017 Received money from other country -0.088 -0.114 Property in CoO -0.307** -0.393** Property in 3rd country -0.405 -0.462 Number of trips to CoO -0.060 -0.039 Number of trips to 3rd country 0.141** 0.090 Voted in last CoO election 0.136 0.049 Consumption of 3rd lang. TV content 0.192 0.153 Shopping abroad 0.104 0.063 Received e-messages from abroad 0.245 0.234 Identification with CoO 0.069 0.097

Constant (cut1) -1.346*** 0.934 1.241* 1.083* 1.154* 1.116* 0.778 0.862 1.286* 1.270* 2.234** Constant (cut2) 0.883*** 3.478*** 3.780*** 3.640*** 3.713*** 3.678*** 3.341*** 3.404*** 3.858*** 3.813*** 4.847*** Observations 1,495 1,329 1,259 1,317 1,322 1,320 1,320 1,318 1,275 1,323 1,184 Pseudo R-squared 0.0187 0.104 0.103 0.106 0.108 0.107 0.106 0.104 0.112 0.103 0.121 *** p<0.001, ** p<0.01, * p<0.05. Additional controls: survey mode (all models); information on CoR language knowledge missing by design (M2-M11); did not live in CoR or was not eligible to vote at last CoO election (M8, M11). EUCROSS, 2016.

The relationship between transnationalism and integration

Groups Based on the literature, this study assumed that the group variables, i.e., the combination of CoO and CoR, would significantly influence the level of individual integration. Table 10 shows that this is not the case for social integration. Only Turkish migrants in Italy differ significantly from their co-nationals in Denmark. Their coefficient indicates that their social circles have a significantly lower probability of including CoR nationals. This effect persists after the introduction of sociodemographic controls, as well as indicators of individual aspects of transnationalism (see M2 to 10). The effect only becomes insignificant in the last model (M11) in which group indicators, sociodemographic variables, and all observed forms of transnationalism are controlled for. Consequently, similar to the situation for transnational practices, the descriptive analysis shows clear differences between the samples in respect to respondents’ social integration in the CoR. However, the regressions show that these differences, for the most part, cannot be explained by circumstances specific to one of the two observed nationalities, particular CoR effects, or a combination of the two. More importantly, the results also show no effect of the status as EU or third country citizen with respect to social integration. To test this further, three additional regressions based on Model 11 were computed (tables not included). Instead of the mentioned group indicators, the first of these models used separate indicators of the CoR and nationality, the second used only indicators of the CoR, and the third used only a dummy variable for nationality. None of these indicators became significant in any of those models.

Sociodemographic and migration-related aspects The second model in Table 10 introduces additional background variables to control for socio- economic status and migration-related aspects, such as age at migration and duration of settlement. All of the following models also control for these aspects.

The different models in Table 10 show that social integration is highly related to the other dimensions of integration differentiated by Esser (2000, 271). The relevance of structural integration manifests itself in the significance of the subjective assessment of the respondents’ household financial situation. The absence of any significant effects of labour market integration indicates that the degree of social integration in the CoR is more strongly influenced by the socio- economic status as such than by the integration in the labour market. However, the directionality of this effect cannot be determined from this result as arguments can be made for both cases: On the one hand, individuals who are financially better off could potentially invest more money in

182

The relationship between transnationalism and integration leisure activities and, by doing so, reinforce and establish social contacts, strengthening their social integration. On the other hand, migrants who have closer social contact with CoR natives might be able to integrate faster structurally due to their access to human capital and knowledge of procedures and mechanisms within the system.

Model 2 also introduces two aspects of cultural integration, namely the self-reported level of knowledge of the CoR language at the time of the survey, and familiarity with one or more regions in the CoR besides the one respondents’ live in. Both indicators of cultural integration are positively related to social integration. However, these effects are hardly surprising: It is safe to assume that migrants who speak the language of their CoR can interact more easily with the general population and, in the process, might find friends amongst those individuals more rapidly. At the same time, it is likely that having a greater number of native speakers in one’s social circle, providing a higher probability of interactions in an informal setting, will enhance migrants’ CoR language capabilities. Hence, while the main effect is likely to go from language knowledge to social networks, both aspects are in fact mutually reinforcing. The relation between broader knowledge of the CoR and social integration is likely to be fairly similar. The more close contacts of CoR origin individuals have, the higher is the probability that they possess knowledge of different regions in the CoR. For instance, such knowledge could be acquired from visiting close friends or family members who have moved away from the respondents’ area to another region. Furthermore, if respondents befriend people who are originally from other parts of the CoR, these contacts might encourage them to visit these places. However, it is also possible that the relation between these aspects works in reverse, given that visiting and becoming acquainted with other cities and towns, also increases the probability of encountering and befriending people who live there. Finally, both aspects of cultural integration can also be interpreted as a sign of a generally positive stance and certain curiosity towards the CoR, which is likely to translate in a higher interest in interacting with natives.

The models in Table 10 also underline that integration is a process which concerns more than one party, as perceived discrimination is negatively related to social integration in all models. That said, a certain degree of caution is advisable when interpreting this finding since the used indicator does not objectively measure occurrences of discrimination, but rather the subjective feeling of being discriminated against. The analysis shows that respondents who at least occasionally felt discriminated in the CoR were less likely to have friends and family of CoR

183

The relationship between transnationalism and integration origin. This effect remains not only significant if other sociodemographic variables are included in the model but also when all measured forms of transnationalism are controlled for (M11).

The last set of background variables introduced in Model 2 concerns the respondents’ social relations besides friends and family members of CoR origin. In line with the findings of Rodríguez-García et al. (2015, 237), the model shows that having a CoR partner is positively related to the number of CoR contacts in the respondents’ social networks. However, the direction of this effect cannot be determined as it is plausible in both ways and might actually vary between respondents. This means that having a CoR partner could lead to a high number of CoR friends, but having a lot of CoR friends could also increase the chances of engaging in a romantic partnership with a CoR native. Furthermore, Model 2 indicates that having a higher number of acquaintances from other backgrounds, be it migrants from the CoO or third countries, is positively related to the degree of social integration, i.e., the number of CoR natives in the respondents’ social networks. However, the positive effect is more consistent for friends and family with a third country background, remaining visible in all models, than for CoO acquaintances where it is not present in some models, most importantly in Model 11.

Transnational background and networks Of the variables introduced in Model 3, only the dummy variable indicating that respondents were separated from their life partner by international frontiers has a negative coefficient. In principal, this is intuitive given that the existence of a partner is generally more likely to expand one’s pool of potential friends and family members, rather than to limit it. However, if the partner is living in another country, such an effect is likely to focus on this country as well. Consequently, no positive effect of having a partner abroad on social networks in the CoR could be expected. More important than such theoretical deliberations, however, is the fact that this coefficient is neither significant in M3 nor in M11, indicating that there is no statistically significant effect of being in a romantic cross-border relationship on social integration in the CoR.71 Furthermore, the indicators of transnational networks all show positive coefficients in M3 and M11. In the case of CoO friends and family in a third country, this effect is even statistically significant in both models (M3 and M11). Therefore, hypothesis HI1, which stated that a negative relation between social integration in the CoR and the integration in transnational social networks

71 A bivariate analysis for the overall sample (not reported) does not show a significant negative effect of having a partner abroad on social integration, either.

184

The relationship between transnationalism and integration should exist, is not supported by these findings. This is a highly important result, which means that social interactions with CoR nationals and social connections across borders cannot be perceived in terms of an either/or relationship. Furthermore, the findings contradict similar predictions, made by Esser (2001, 21), with respect to the relation between the integration in different networks within the CoR. According to this data, multiple integration in different social groups is indeed possible. In fact, it is safe to assume that the individuals of the above- differentiated backgrounds do in many cases not form distinct networks but are often part of diverse social circles. Accordingly, the fact that significant positive effects can be observed indicates that, with respect to social integration in the CoR, the inclination of migrants to socialise with others is more important than the location or ethnicity of the individuals they socialise with.

Transnational human capital Models 4 and 11 show that none of the indicators of transnational human capital is significantly associated with migrants’ social integration in the CoR. This indicates that these aspects neither hinder nor reinforce social integration.

Transnational communication Model 5 shows the effects of the three measured means of transnational communication. Of those indicators, one (communication via mail and email) is significantly related to social integration. However, this effect is no longer significant in the full model (M11), i.e., when all the other measured aspects of transnationalism are controlled for. These findings suggest that respondents who frequently interact with others across borders are not socially less integrated in their CoR. Consequently, the overall model supports hypothesis HI4, according to which no significant relation exists between transnational communication and social integration.

Economic transnationalism Importantly, Model 6 stresses that there are indeed transnational practices which are negatively and significantly related to social integration. More precisely, this is true for the existence of property in the CoO. A causal relation between this aspect and social integration is possible but not imperative. It is conceivable that migrants who do not find access to relevant social circles in the CoR could be more likely to envision their return to the CoO and, therefore, buy property in the respective country. Furthermore, it is possible that people who possess such property prior to

185

The relationship between transnationalism and integration their migration could be more likely to see their life in the CoR as a temporary endeavour right from the start, and consequently put less effort into establishing meaningful relations with natives in the CoR. However, neither of these theoretical explanations for this negative effect can be verified by the data at hand. The negative relation between real estate ownership in the CoO and social integration also persists if the model controls for other forms of transnationalism (M11). Hence, this is one of only two indicators of transnationalism which are significant in the complete model. In contrast to longstanding (financial) commitment to property in the CoO, the fact and frequency of sending remittances is not negatively related to social integration in itself. This is quite relevant, given that the remitting of money is – partly due to its economic implications for the CoO – an often discussed topic in migration and integration research. A bivariate analysis (table not included) does also show no significant relation between remitting behaviour and social integration. Nevertheless, as a negative relation between an indicator of economic transnationalism (i.e., property ownership in the CoO) and social integration is visible, HI3, which stipulated that no relation would exist between this type of practice and the dependent variable, has to be rejected with regard to social integration.

Transnational mobility Model 11 shows that HI5 is supported by the present data, as no significant effect of transnational mobility on social integration can be observed in the full model. This is especially interesting in light of the reported negative relation between property in the CoO and social integration. It means that the mere existence of such property is more relevant than the question of how often the respondents travel to the CoO and use their property personally. A positive effect of the number of visits to third countries is only significant as long as the model does not control for other indicators of transnationalism (i.e., in M7).

Political transnationalism, transnational consumption and identification with the CoO Models 8 to 10 and Model 11 show that neither political transnationalism, transnational consumption, nor the degree of identification with the CoO are significantly related to social integration.

After reviewing the relation between transnationalism and social integration, the following subchapter will present similar analyses with respect to identificational integration on a national

186

The relationship between transnationalism and integration and town level. A summary and conclusions for all three sets of regressions are provided at the end of the chapter.

6.2 Identificational integration This subchapter takes a closer look at the relation between transnationalism and identificational integration at the level of the country and town of residence. As described in chapter 3.4, identification was measured by the degree to which respondents agreed with statements like “I feel Danish/German/Italian”. Table 11 provides information regarding respondents’ identification with the town and country they live in as well as with their country of origin. Besides the mean values, the table also includes the distribution of answers across the used five- point scale. These values are given as percentages of each group.

Table 11: Identification with country of residence, town of residence, and country of origin

Distribution in percent Strong Strong Mean disagreement agreement (1) (2) (3) (4) (5)

Indentification with the CoR

TU in DK 1.7 67 9 15 5 4 TU in DE 1.3 82 7 9 2 0 TU in IT 2.0 55 13 14 11 6 RO in DK 2.1 46 17 19 14 4 RO in DE 2.8 35 8 19 16 23 RO in IT 2.4 43 9 21 16 11

Indentification with the ToR

TU in DK 3.4 23 7 17 13 40 TU in DE 3.4 20 7 21 17 35 TU in IT 3.1 26 4 25 21 24 RO in DK 3.4 10 11 29 28 22 RO in DE 3.7 13 9 18 18 42 RO in IT 3.8 11 7 18 24 41

Indentification with the CoO

TU in DK 4.5 5 4 8 6 77 TU in DE 4.7 4 1 5 4 86 TU in IT 4.4 5 2 11 11 71 RO in DK 4.4 5 4 9 13 68 RO in DE 4.3 5 5 14 11 65 RO in IT 4.5 3 4 9 11 72 EUCROSS, 2016. N = 1,475.

187

The relationship between transnationalism and integration

Table 11 shows that all samples score low on the indicator used to measure CoR identification. However, it is important to take the study design into account when interpreting this result: As laid out in chapter 3.3, only migrants who did not hold the nationality of the CoR at the time of the survey were eligible to participate in it. Unfortunately, the survey did not collect any data on whether respondents did not want to get naturalised, or were prevented from doing so by structural or legal barriers. In any case, it is possible that the respondents differ systematically in their assessment of the statement “I feel Danish/German/Italian” from migrants who did get naturalised in their CoR, and, consequently, could not participate in the survey this analysis is built upon.

Across all the Turkish subsamples, the majority of respondents strongly disagree with the mentioned statement. However, the distribution of answers across the scale indicates some degree of variation. Most importantly, it becomes obvious that the identification with the CoR is weakest amongst Turkish respondents in Germany, where 82 percent of the sample indicated strong disagreement with this statement. In this regard, it is noteworthy that Romanians in Germany show, on the contrary, the highest level of CoR identification according to the mean values. Romanians in Denmark and Italy also showed a higher level of CoR attachment than their Turkish counterparts. This means that the level of identification with the country that respondents live in is noticeably higher amongst the surveyed EU citizens than amongst migrants with third country status. Hence, while no clear pattern of higher or lower identification is visible based on the country in which respondents reside, the tendency with regard to their nationality is consistent across all three CoRs.

While respondents in all samples show a tendency not to identify with the CoR, the opposite is true with respect to the CoO. In each of the six samples, the overwhelming majority of participants show the highest possible level of agreement with the respective statement. If the mean values are taken as criteria, few differences between Turkish and Romanian migrants are visible.

The level of identification with respondents’ town of residence takes a middle ground. For all samples, it is on average considerably higher than CoR identification. It becomes apparent that, with exception of the Turkish migrants in Italy, 50 percent or more of the respondents across all samples identify to some extent with the town that they live in (i.e., chose four or five on the

188

The relationship between transnationalism and integration scale). The finding of higher identification with the ToR rather than the CoR is in line with the results for direct descendants of migrants reported by Schneider and co-authors (2012, 312-13). A somewhat different instrument is used by Bochove et al. (2010, 356), who asked their sample of middle-class migrants in Rotterdam to indicate their degree of connection with co-nationals in the Netherlands, in the CoO, and in third countries, as well as with the citizens of Rotterdam and native Dutch people in general. On the one hand, the authors found – similar to the present case – that their respondents relate more strongly to the citizens of Rotterdam than to the native population. On the other hand, they found the same hierarchy between local attachment and connections with co-nationals in the CoO, with the latter being somewhat weaker (2010, 356). In both texts, the authors argue that local identification might function in some regard as an alternative to identification with the CoR. This could be due to a situation in which the differentiation between in and out-groups, in which migrants are styled as ‘the other’ by politicians, media, and the native population, is to a stronger degree present at the national rather than the local level (Schneider et al. 2012, 312-13). However, it is also conceivable that migrants identify more strongly with their ToRs because these are either perceived to be more multi- cultural than the CoR as a whole, or because a high number of other migrants, if not even co- nationals, are living in their direct neighbourhood, effectively transforming the character of this space, to the end that it might be perceived to be essentially Turkish or Romanian by respondents (Ehrkamp 2005; Bochove, Rusinovic, and Engbersen 2010, 356). Unfortunately, the used data does not allow controlling for the composition of the respondents’ residential area. Besides the possible explanations derived from empirical work, a theoretical argument based on Lawler (1992) (see chapter 2.4) could be made that migrants might identify stronger with the ToR as it may simply be more present and relevant in respondents’ daily lives compared to the somewhat more abstract idea of the CoR.

In any case, the discussed differences between these two aspects of identification suggest that identificational integration with respect to country and town of residence occur, at least in part, independently from each other and are not determined in the same manner by an equal set of influencing factors. The following regressions will investigate in more detail how both forms of identificational integration are related to transnationalism and influenced by sociodemographic aspects.

189

The relationship between transnationalism and integration

6.2.1 Identificational integration in the country of residence Table 12 presents linear regression models which use the degree of identification with the CoR as a dependent variable. The sequence of models is similar to that presented for social integration in Table 10: the first model includes only group indicators, while subsequent models gradually introduce socio-economic and migration-related background variables, and indicators of transnationalism and identification with the CoO. Finally, a complete model is presented.

The independent variables used in the following regressions are described in Appendix 6. For an overview of all hypotheses see Table 1.

190

Table 12: Linear regression of identification with the country of residence (coefficients) M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Groups TU in DK (ref.) TU in DE -0.375** -0.394** -0.392** -0.416** -0.334** -0.394** -0.373** -0.391** -0.410** -0.372** -0.348* TU in IT 0.256 0.278 0.242 0.299 0.284 0.341 0.274 0.288 0.350 0.315 0.445 RO in DK 0.446*** 0.599*** 0.561*** 0.645*** 0.631*** 0.589*** 0.613*** 0.584*** 0.632*** 0.597*** 0.619*** RO in DE 1.123*** 0.881*** 0.884*** 0.902*** 0.924*** 0.910*** 0.894*** 0.866*** 0.910*** 0.866*** 0.978*** RO in IT 0.734*** 0.780*** 0.777*** 0.785*** 0.815*** 0.807*** 0.814*** 0.799*** 0.821*** 0.780*** 0.876*** Gender (female = 1) 0.038 -0.001 0.021 0.037 0.043 0.039 0.030 0.059 0.059 0.063 Married/ registered civil union (ref.) Single, never been married -0.134 -0.138 -0.122 -0.135 -0.169 -0.132 -0.161 -0.140 -0.188 -0.243* Other 0.243* 0.246 0.265* 0.238 0.227 0.235 0.238 0.180 0.208 0.132 Age at migration 0.002 0.000 0.002 0.003 0.005 0.001 0.003 0.001 0.001 0.003 Duration of stay 0.025*** 0.023*** 0.025*** 0.025*** 0.027*** 0.024*** 0.025*** 0.027*** 0.024*** 0.028*** Education Lower secondary or less (ref.) In-between low. and higher second. -0.150 -0.148 -0.112 -0.185 -0.124 -0.144 -0.133 -0.192 -0.108 -0.093 Higher secondary 0.077 0.081 0.095 0.060 0.081 0.070 0.091 0.012 0.096 0.072 Tertiary 0.348** 0.355** 0.381** 0.305** 0.336** 0.365** 0.372*** 0.284* 0.355** 0.333** Received highest education in CoR 0.271** 0.300** 0.286** 0.272** 0.276** 0.258** 0.301** 0.240* 0.233* 0.252* Labour market participation Working (ref.) Non-Working 0.010 0.015 0.013 0.019 -0.048 0.013 0.019 0.035 0.013 -0.001 Retired -0.000 0.001 -0.033 0.021 -0.076 0.010 0.002 0.007 0.017 -0.065 Unemployed 0.057 0.040 0.037 0.051 0.019 0.054 0.063 0.053 0.068 0.001 Other 0.006 0.002 0.031 0.021 -0.049 -0.003 0.015 -0.087 -0.003 -0.131 Household financial sit. at age 14 -0.078* -0.085* -0.081* -0.073* -0.080* -0.083* -0.072 -0.083* -0.077* -0.096* Household financial sit. now 0.184*** 0.204*** 0.195*** 0.191*** 0.211*** 0.187*** 0.194*** 0.197*** 0.185*** 0.258*** Property in the CoR -0.038 -0.080 -0.030 -0.048 -0.025 -0.058 -0.034 -0.069 -0.046 -0.106 Discrimination experience -0.273*** -0.265*** -0.276*** -0.272*** -0.261*** -0.273*** -0.257*** -0.258*** -0.268*** -0.236** Knowledge of other CoR regions 0.027 0.027 0.040 0.028 0.032 0.025 0.029 0.015 0.026 0.041 Knowledge of the CoR language No or poor knowledge (ref.) Intermediate knowledge 0.133 0.096 0.113 0.125 0.129 0.135 0.128 0.075 0.136 0.029 Good or very good knowledge 0.360* 0.328* 0.371** 0.336* 0.344* 0.361* 0.340* 0.293* 0.344* 0.212 CoR partner 0.137 0.079 0.143 0.126 0.136 0.151 0.123 0.136 0.114 0.096 Social contacts in the CoR from CoO -0.120* -0.127* -0.127* -0.131* -0.101 -0.114* -0.116* -0.102 -0.109* -0.087 from 3rd country 0.087 0.089 0.097 0.080 0.086 0.072 0.077 0.059 0.074 0.047

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Social contacts in other countries from CoO in CoO -0.041 -0.019 from CoO in 3rd country -0.005 0.017 from 3rd country living abroad -0.018 -0.036 Partner lives in other country -0.313 -0.204 Knowledge of a 3rd language -0.111 -0.112 (at least moderate) Knowledge of 3rd countries -0.045 -0.083 Previous migration experience -0.131 -0.097 Frequency of transnat. communication via phone and VoIP -0.125** -0.083 via mail and email 0.021 0.016 via social netw. sites 0.055* 0.035 Frequency of remittances -0.041 -0.028 Received money from other country 0.099 0.101 Property in CoO -0.265*** -0.240** Property in 3rd country -0.688* -0.756* Number of trips to CoO -0.073* -0.030 Number of trips to 3rd country 0.031 0.027 Voted in last CoO election -0.078 -0.000 Consumption of 3rd lang. TV content 0.051 0.083 Shopping abroad -0.083 -0.037 Received e-messages from abroad 0.213** 0.172 Identification with CoO -0.153*** -0.148***

Constant 1.699*** 0.631 0.794* 0.692* 0.822* 0.547 0.728* 0.535 0.535 1.343*** 1.395*** Observations 1,489 1,326 1,256 1,314 1,319 1,317 1,317 1,315 1,271 1,322 1,183 Adj. R-squared 0.123 0.208 0.208 0.211 0.213 0.219 0.210 0.208 0.209 0.223 0.237 *** p<0.001, ** p<0.01, * p<0.05. Additional controls: survey mode (all models); information on CoR language knowledge missing by design (M2-M11); did not live in CoR or was not eligible to vote at last CoO election (M8, M11). EUCROSS, 2016.

The relationship between transnationalism and integration

Groups Contrary to the social integration results, Table 12 shows significant effects of the group identifiers on respondents’ identification with their CoR. Furthermore, these effects are persistent throughout all models, i.e., they remain significant even if sociodemographic aspects and transnationalism are controlled for (M11). These results indicate that Turkish migrants in Italy do not differ significantly in their CoR identification from their co-nationals in Denmark, while Turks in Germany are significantly less likely to identify with their CoR than respondents of the reference group. The indicators of all three Romanian migrant groups are highly significant, and their positive coefficients suggest that the respondents in these groups are more likely to identify with their CoR than Turkish migrants in all three countries. The fact that coefficients, especially in the full model (M11) are more similar when viewed by nationality, rather than by CoR, suggests that differences are more strongly driven by the origin of respondents than by the settlement contexts. This assessment is supported by three additional regressions (tables not included) which used Model 11 but replaced the group identifiers: In the first regression, a dummy variable for the CoO was used alongside identifiers for the CoR. In this model, Romanian nationality was highly significant, while none of the CoR variables were significant (adjusted R² 0.226). The second regression only used the aforementioned dummy variable that differentiated between respondents nationalities without controlling for the CoR. The coefficient was again highly significant and indicated that Romanians are more likely to identify with the country they live in than Turks (adjusted R² 0.226). Finally, the third regression used the variables that identified the respondents’ CoR without differentiating between the two nationalities. This model only showed a significant difference for Denmark and Italy, with respondents in the latter country being significantly more likely to identify with the CoR (adjusted R² 0.191). As mentioned, these results support the initial assessment according to which the group differences in Table 12 indicate rather an effect of the respondents’ CoO than of contextual factors specific to the CoRs or a combination thereof. At the same time, the group effects cannot be attributed to stronger identification with the CoO since the model includes a separate control for this (see below). This means that the observed group effects are unlikely to be related to possible nationalist or ethnocentric sentiments, as these are more likely to have been captured by the additional indicator of CoO identification. Consequently, it can be assumed that the described group effect is to a considerable degree due to differences in legal status that both groups experience in all three countries under investigation. These legal circumstances are – in direct comparison – more

193

The relationship between transnationalism and integration favourable for Romanians as intra-EU migrants than for migrants from Turkey. Besides any practical consequences, it is possible that there also is an affective component: Romanians might feel closer to their CoR than Turks, due to the fact that it belongs to the same supranational organisation as their CoO, which means that the status as EU citizens is something Romanian migrants share with the native population. In short, while not possessing CoR citizenship, the Romanians’ status as European citizens puts them in a privileged position compared to migrants from third countries and may well have a positive impact on their feelings towards the country they live in. These findings support hypothesis HI6, which assumed that Romanian migrants would show a stronger degree of identification with the CoR than Turkish migrants.

Sociodemographic and migration-related aspects Model 2 shows the significance of several socio-economic indicators which persist – in most cases – throughout the following models. Amongst those are positive effects of tertiary education and of having obtained the highest formal education in the CoR. The latter means that socialisation within a country’s educational system favours the identification with the same country.

Another aspect which shows that identificational integration is related to structural integration is the strong effect of a positive evaluation of the respondents’ current household situation. Additionally, both M2 and M11, show that the assessment of the household financial situation the respondents lived in at age 14 is negatively associated with their identification with the CoR. This means, the more difficult the financial situation of a household was, the stronger the respondents’ identificational integration. Taken together, both indicators measuring a household’s past and present financial situation imply that respondents who consider themselves to be living in a more favourable financial situation nowadays, compared to when they were adolescents, are more likely to identify with the CoR. Furthermore, the data show that only around ten percent of the overall sample was already living in their CoR before the age of 15. This means the assessment of the situation at age 14 mostly refers to pre-migratory households. This suggests that respondents are more likely to identify with their CoR if they judge their migratory project to have been overall successful in improving their living conditions.

Interestingly, in addition to being educated in the CoR, the duration of stay in the respective country has a significant positive effect on identification with it, i.e., the more time respondents

194

The relationship between transnationalism and integration spent in a country, the more likely they are to feel a sense of belonging.72 This corresponds to the findings of Snel and co-authors (2006, 302). Furthermore, a negative impact of discrimination experience can be observed. Both effects are consistent with the findings of Braun and Müller (2012, 280) regarding CoR attachment of intra-EU migrants. However, it is interesting that the negative effect of discrimination on CoR identification does not translate to a similar positive effect on identification with the CoO. Additional regressions (tables not included) show that discrimination experiences are not significantly related to CoO identification in a bivariate analysis nor in a model which controls for the same background variables as M2 above. In other words, this means that discrimination is associated with a lower degree of CoR identification but not with a higher degree of CoO identification.

Identificational integration at the CoR level does not appear to be strongly related to cultural integration. This becomes apparent as broader personal knowledge of the country respondents live in is not significant in any of the models. Additionally, a good or very good level of the CoR language is significant in M2, but this effect is not present in M11 anymore.

Regarding social contacts, a negative effect of friends and family from the CoO can be observed in Model 2 but not in Model 11. In fact, Models 6 and Model 9 indicate that this effect loses its significance as soon as either economic transnationalism or transnational consumption is controlled for. In particular, the influence of property in the CoO (as an indicator of economic transnationalism) is of interest as this practice signals a connection between migrants and their CoO. The same holds true for the network indicator to some extent. Both coefficients (property in the CoO and CoO friends and family in the CoR) are negative, which means that their effects do not cancel each other out. Since a similar situation presents itself with regard to visits to the CoO the implications of this result will be discussed in more detail below.

Transnational background and networks, transnational human capital Neither transnational background and networks (M3) nor transnational human capital (M4) appear to be significantly related to the degree of identification with the CoR.

72 Additional computations (tables not included) confirm this effect. The duration of stay remains significant in M11 when combined with age (which is not significant) instead of age at migration. If age at migration and age are used instead, the coefficient of the former is negative and the coefficient of the latter is positive, with both effects being highly significant. This, again, indicates that it is, in fact, the duration of respondents’ stay in the CoR (i.e., the difference of the other two indicators) which is strongly related with the dependent variable.

195

The relationship between transnationalism and integration

Transnational communication With regard to transnational communication, two different effects can be observed in Model 5. While the frequency of communication via social networking sites shows a positive relation with the level of CoR identification, direct verbal transnational communication via phone or VoIP is negatively related to it. This indicates that the effect of communication differs depending on its characteristics. However, it is not clear if the relevant characteristics lie on the side of the users, or in the nature of the means of communication itself. In any case, Model 11 shows that both effects are no longer significant if all other measured forms of transnationalism are controlled for. This model consequently lends support to hypothesis HI4, indicating the absence of a significant association between transnational communication and identificational integration at the CoR level.

Economic transnationalism and transnational mobility Model 6 reveals that property in a third country and property in the CoO are negatively related with CoR identification. However, purely monetary forms of economic transnationalism show no significant impact on the dependent variable. In contrast, the negative effects of the two real estate variables are also visible in the full model (M11). This means that hypothesis HI3, which assumed that no significant relation between economic transnationalism and identificational integration would exist, is not supported by the present findings.

A significant negative effect of transnational mobility to the CoO is only visible in Model 7, while the effect is not significant in the overall model (M11). Mobility to third countries does not show a significant relation with identification at all. These findings support hypothesis HI5, which assumed the absence of any relation between the three types of integration and transnational mobility. Furthermore, an additional regression (table not included) shows that the indicator of visits to the CoO becomes insignificant as soon as property in the CoO is controlled for.73 As mentioned above, the same holds true for the negative effect of a high number of CoO friends and family members in the CoR. Taken together, all three aspects indicate that substantial connections to the CoO and interactions with co-nationals can have a negative impact on CoR identification. However, the two weaker effects (mobility and social relations) are mediated by the effect of real estate ownership which also indicates rather a long-term financial investment.

73 However, the indicator remains significant if transnational consumption is controlled for instead of property in the CoO.

196

The relationship between transnationalism and integration

This means this type of substantial connection – which also might be a sign that the idea of returning to the CoO is kept alive – is more strongly and negatively related to integration, than the sheer existence of day to day relations with other CoO nationals, or the frequency of visits to the country migrants originated from.

The absence of significant effects of transnational mobility and communication in the final model is also interesting because Snel et al. (2006) report in their analysis a negative impact of social- cultural transnationalism on identification with the CoR population. However, they, unfortunately, use a compound indicator combining visits to family and friends in the CoO, frequent contact with them, and membership in CoO organisations. Therefore, it is not possible to judge which of the included aspects may have driven the effect reported by them and whether the findings of this study support their results.

Political transnationalism Model 8 and 11 show that, as in the case of social integration, political transnationalism in the form of absentee voting is not significantly related to CoR identification.

Transnational consumption Rather surprisingly, receiving electronic messages from abroad during the 12 months preceding the survey is significant in Model 9. This effect is somewhat puzzling. Additional computations (tables not included) show that the effect remains significant in models which combine transnational consumption with nearly all of the other types of transnationalism, and with the identification with the CoO. The only exception is transnational communication. Hence, the indicator of electronic messages loses its significance if all three forms of transnational communication are added to Model 9. Furthermore, it is significant in M11 if these indicators are deleted from the model. This indicates that the positive effect of this indicator, which, in principle, suggests a certain degree of transnational communicative connectivity, is superimposed by the effects of more specific communication behaviour. In the full model, however, neither this aspect of transnational consumption, nor transnational communication are significant anymore.

Identification with the CoO Finally, Model 10 shows a highly significant negative relation between CoO and CoR identification. This effect also remains in M11 which controls for all observed aspects of transnationalism in addition to the socio-economic aspects. This indicates that the identifications

197

The relationship between transnationalism and integration with these two entities are, at least to some degree, at odds with each other. Consequently, hypothesis HI2 is supported by the data. Hence, these results differ from those of Snel et al. (2006, 302) who did not find a direct connection between CoO and CoR identification. This difference might partially be related to the nature of the present sample. As described, the recruitment procedure explicitly targeted migrants who did not become naturalised in their CoR and, in contrast to the study of Snel and co-authors, excluded the children of immigrants who were born in the CoR. Individuals in this group could, by default, be expected to identify more strongly with their CoO as this is the country that they were born and socialised in, and of which they hold the nationality. As a consequence, their emotional and substantial connection to the CoO is likely to differ from that of migrants who took on another (or a second) citizenship, and of the children of migrants who grew up in the CoR and who might mostly feel a symbolic connection to the origin country of their parents. In this context, it is worthwhile to reiterate that the negative effect of discrimination on identification with the CoR is not mirrored by a positive relation of discrimination experiences on identification with the CoO. Furthermore, the above- mentioned additional regression – which corresponds to M2 but uses the identification with the CoO as dependent variable – does not show any significant effects of structural integration (e.g., labour force status or household financial situation). This means that the negative association of CoO identification with a feeling of belongingness to the CoR cannot be interpreted as sign of what Portes (1999, 466) has called ‘reactive ethnicity’, i.e., a situation in which a feeling of personal rejection or unfair exclusion in the country of settlement leads to a stronger identification with the country and culture of origin.

6.2.2 Identificational integration in the town of residence In the following, Table 13 presents regression models similar to those in Tables 10 and 12 but uses the respondents’ identification with the ToR as dependent variable.

The independent variables used in the following regressions are described in Appendix 6. For an overview of all hypotheses see Table 1.

198

Table 13: Linear regression of identification with the town of residence (coefficients) M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Groups TU in DK (ref.) TU in DE 0.023 -0.025 -0.040 -0.023 0.034 -0.032 -0.013 -0.030 0.061 -0.028 0.070 TU in IT 0.085 -0.119 -0.180 -0.114 -0.116 -0.031 -0.118 -0.111 -0.009 -0.124 -0.026 RO in DK 0.029 0.220 0.182 0.251 0.260 0.212 0.215 0.230 0.227 0.212 0.239 RO in DE 0.302* 0.051 0.052 0.061 0.108 0.073 0.043 0.042 0.115 0.057 0.179 RO in IT 0.401** 0.373* 0.372* 0.373* 0.414* 0.379* 0.383* 0.395* 0.473** 0.368* 0.534** Gender (female = 1) -0.018 -0.047 -0.028 -0.016 -0.012 -0.018 -0.024 -0.006 -0.023 -0.013 Married/ registered civil union (ref.) Single, never been married -0.094 -0.070 -0.098 -0.095 -0.101 -0.085 -0.105 -0.110 -0.090 -0.096 Other -0.136 -0.119 -0.116 -0.148 -0.140 -0.132 -0.145 -0.177 -0.126 -0.154 Age at migration 0.009 0.008 0.008 0.010 0.010 0.009 0.010 0.010 0.009 0.011 Duration of stay 0.025*** 0.024*** 0.024*** 0.026*** 0.026*** 0.025*** 0.026*** 0.028*** 0.025*** 0.029*** Education Lower secondary or less (ref.) In-between low. and higher second. -0.401* -0.367* -0.412* -0.445** -0.377* -0.392* -0.374* -0.366* -0.405* -0.382* Higher secondary -0.096 -0.087 -0.089 -0.120 -0.080 -0.099 -0.079 -0.156 -0.102 -0.126 Tertiary 0.044 0.051 0.045 -0.017 0.046 0.045 0.066 -0.056 0.047 -0.049 Received highest education in CoR 0.245* 0.226* 0.254* 0.237* 0.240* 0.225* 0.264* 0.185 0.246* 0.189 Labour market participation Working (ref.) Non-Working 0.022 0.013 0.023 0.035 -0.014 0.026 0.041 0.065 0.026 0.033 Retired -0.099 -0.096 -0.091 -0.073 -0.153 -0.083 -0.097 -0.030 -0.094 -0.068 Unemployed -0.088 -0.106 -0.109 -0.094 -0.108 -0.086 -0.078 -0.049 -0.080 -0.088 Other -0.033 -0.061 -0.020 -0.018 -0.065 -0.022 -0.027 -0.025 -0.029 -0.079 Household financial sit. at age 14 -0.092* -0.085* -0.091* -0.091* -0.093* -0.096* -0.090* -0.081 -0.095* -0.078 Household financial sit. now 0.159** 0.146** 0.157** 0.164** 0.176*** 0.156** 0.168** 0.169** 0.160** 0.186*** Property in the CoR 0.054 0.008 0.060 0.041 0.079 0.034 0.046 0.032 0.061 -0.007 Discrimination experience -0.503*** -0.475*** -0.504*** -0.502*** -0.501*** -0.497*** -0.496*** -0.478*** -0.501*** -0.445*** Knowledge of other CoR regions 0.029 0.035 0.024 0.030 0.033 0.025 0.027 0.020 0.031 0.017 Knowledge of the CoR language No or poor knowledge (ref.) Intermediate knowledge 0.207 0.158 0.188 0.195 0.208 0.199 0.198 0.180 0.205 0.064 Good or very good knowledge 0.558*** 0.563*** 0.556*** 0.514** 0.543*** 0.554*** 0.538** 0.471** 0.556*** 0.385* CoR partner 0.145 0.080 0.114 0.121 0.145 0.146 0.137 0.145 0.150 0.057 Social contacts in the CoR from CoO -0.027 -0.050 -0.037 -0.045 -0.014 -0.021 -0.017 -0.024 -0.030 -0.034 from 3rd country 0.100 0.102 0.105 0.090 0.102 0.083 0.095 0.080 0.109 0.082

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 Social contacts in other countries from CoO in CoO -0.002 0.015 from CoO in 3rd country 0.045 0.040 from 3rd country living abroad -0.056 -0.094 Partner lives in other country -0.913*** -0.875*** Knowledge of a 3rd language -0.046 -0.069 (at least moderate) Knowledge of 3rd countries 0.033 0.023 Previous migration experience -0.079 -0.041 Frequency of transnat. communication via phone and VoIP -0.146** -0.106* via mail and email 0.049 0.000 via social netw. sites 0.076* 0.049 Frequency of remittances -0.049 -0.042 Received money from other country 0.009 0.027 Property in CoO -0.115 -0.109 Property in 3rd country -0.610 -0.545 Number of trips to CoO -0.020 0.003 Number of trips to 3rd country 0.042 0.031 Voted in last CoO election -0.147 -0.125 Consumption of 3rd lang. TV content 0.154 0.185* Shopping abroad -0.025 -0.006 Received e-messages from abroad 0.316*** 0.321** Identification with CoO 0.032 0.040

Constant 3.393*** 2.385*** 2.521*** 2.437*** 2.590*** 2.349*** 2.393*** 2.301*** 2.035*** 2.251*** 2.081*** Observations 1,489 1,325 1,254 1,313 1,318 1,316 1,317 1,314 1,270 1,320 1,181 Adj. R-squared 0.020 0.116 0.123 0.114 0.124 0.117 0.115 0.114 0.118 0.115 0.127

*** p<0.001, ** p<0.01, * p<0.05. Additional controls: survey mode (all models); information on CoR language knowledge missing by design (M2-M11); did not live in CoR or was not eligible to vote at last CoO election (M8, M11). EUCROSS, 2016.

The relationship between transnationalism and integration

Groups The first rows in Table 13 show that, contrary to CoR identification, differences between migrant groups are of little importance regarding the identification with the ToR. Only those Romanians who live in a country other than the reference group (Turkish migrants in Denmark) show a significantly stronger identification with the ToR in the first Model. However, the dummy variable for Romanians in Germany becomes insignificant already in Model 2, i.e., upon the introduction of controls for sociodemographic and migration-related aspects. As in the previous case (i.e., for identificational integration at CoR level), additional versions of M11 were computed to test for the effects of CoR and CoO separately (tables not included). These computations showed that differences between the two nationalities were only visible if the model did not simultaneously control for the CoR. It was only in this instance that the Romanians were significantly more likely to identify with the town they resided in compared to the Turks. Consequently, there is some indication that the intra-EU migrants identify more strongly with their ToR than the third country migrants. If the model controlled only for the CoRs – i.e., without differentiating between the two nationalities – the indicator of migrants in Italy was significant, indicating that these respondents are somewhat more likely to identify with their ToR than those in Denmark.

Sociodemographic and migration-related aspects The second model in Table 13 shows that local identification is not significantly related to the individuals’ labour market participation, but to a positive evaluation of the household financial situation. Contrary to identification with the CoR, having received the highest educational title in the CoR is not significant in the full model (M11).

The effects of higher educational titles are quite interesting: The significance of the negative coefficient of a title in-between lower and higher secondary education in all models infers that these respondents were less likely to identify with the ToR than the individuals in the reference group. Thus in regard to this dimension, higher education does not only lack a positive relation with integration but even displays a negative affiliation with it. However, while all higher educational titles have negative coefficients in M11, only the mentioned one (i.e., in-between lower and higher secondary education) is significant.

The relation between identificational integration at a local level and cultural integration appears to be smaller than in the case of social integration, yet stronger than with respect to identification with the CoR. As in the former case, a comparatively high level of CoR

201

The relationship between transnationalism and integration language knowledge is positively related to identification with the ToR, while a broader knowledge of the CoR itself does not have a significant effect.

Models 2 and 11 also attest to the fact that the identification with the town that migrants settled in is positively related to the duration of their presence in the CoR.74 Furthermore, it is again apparent from the computed models that perceived discrimination is negatively related to integration.

Transnational background and networks The three general indicators of transnational social networks (i.e., those built based on the questions regarding origin and location of friends and family) show no significant associations with ToR identification. However, a highly significant negative effect of having a partner (i.e., husband, wife, boyfriend etc.) in another country is visible in Models 3 and 11. Consequently, this effect is not so much related to the general sociability of respondents, but rather their life circumstances and that of their most immediate family members. The effect could in this case, again, work in both directions. On the one hand, individuals whose family life is not centred in their ToR could display a lower identification with the latter. This could be due to the fact that a substantial part of their most personal social life takes part transnationally. On the other hand, if a higher level of identification is interpreted as an indicator of general satisfaction with their living circumstances in that town, a lower level of satisfaction might negatively influence the probability of relocating ones’ family permanently.

Transnational human capital As in the case of identification with the CoR, transnational human capital does not significantly influence respondents’ sense of belonging to the town they live in.

Transnational communication Model 5 shows that transnational verbal communication has a negative and strongly significant association with the dependent variable, while communication via social networking sites (SNS) shows a positive relation. This corresponds to the previously observed relations between these variables and CoR identification in Model 5 of Table 12. However, while, in this case, none of the effects persisted in the full model, the negative effect of communication via phone and VoIP is still significant in M11 of the identificational integration at a local level. At first glance, it could be hypothesised that these observed effects are indicative of an underlying age effect. This could be the case if it were assumed that

74 Additional computations (tables not included) confirm the effect of the duration of stay in the same way as in the case of identificational integration in the CoR (see footnote 72).

202

The relationship between transnationalism and integration younger respondents would be more likely to use SNS, while older individuals would prefer telephone conversations. However, if age is used instead of age at migration in M5 and M11 these effects remain unchanged (tables not included). Age, on the contrary, does not become significant. In consequence, HI4, which assumed the general absence of a relation between transnational communication and integration, is not supported in respect to this dimension. However, the fact that only one of the three indicators of transnational communication has a negative impact on the dependent variable suggests that this relation cannot be generalised to communication and its frequency, but that it is related to features specific to interactions via telephone or VoIP.

Economic transnationalism, transnational mobility and political transnationalism Models 6, 7, 8 and 11 in Table 13 show the absence of any significant relation between identification with the ToR and economic transnational practices, transnational mobility, as well as political transnationalism. These findings lend support to HI3 and HI5 which assumed the absence of effects of economic transnationalism and transnational mobility.

Transnational consumption The full model (M11) shows that transnational consumption – in the form of the two media usage related practices – is positively associated with the level of identification with the town that respondents live in. Shopping behaviour, however, shows no effect. The results presented in chapter 5 suggested that younger respondents might be more likely to engage in transnational consumption. Therefore, additional models were computed (tables not included) to test whether the positive effects of transnational consumption could indicate an underlying positive association between young age and local identification. However, the results show that this is not the case.

Identification with the CoO Models 10 and 11 show no direct relation – be it positive or negative – between the identification with the town respondents live in and their identification as Turks or Romanians. This finding is remarkable in light of the previously reported results on respondents’ identification with the CoR. If the image of nested identities outlined by Lawler (1992) and Díez Medrano and Gutiérrez (2001) is used, according to which the identification with smaller groups is nested in the identification with larger groups, the findings of this study indicate that a combination of identifications seems to develop more easily if the respective identifications are referring to groups or social structures which are situated on different levels of an imagined continuum.

203

The relationship between transnationalism and integration

6.3 Comparison and summary A compact overview of the variables which are (and which are not) significantly related to any of the three aspects of integration is given in Table 14. This table jointly presents the results of the second model (socio-economic and migration-related variables, M2) and the full model (M11), which combines the former with all indicators of transnationalism and identification with the CoO. The table does not include coefficients but highlights significant variables by indicating their significance level and whether the respective coefficient is of a positive or negative nature.

Table 14: Overview of effects on social and identificational integration Sociodemographic and migration- Full model (M11) related variables (M2)

Social Identific. Identific. Social Identific. Identific. integration with CoR with ToR integration with CoR with ToR Groups TU in DK (ref.) TU in DE – – – TU in IT – RO in DK + + + + + + RO in DE + + + + + + RO in IT + + + + + + + + + Gender (female = 1) Married/ registered civil union (ref.) Single, never been married – Other + Age at migration Duration of stay + + + + + + + + + + + + Education Lower secondary or less (ref.) In-between low. and higher – – second. Higher secondary Tertiary + + + + Received highest education in + + + + CoR Labour market participation Working (ref.) Non-Working Retired Unemployed Other Household financial sit. at age 14 – – – Household financial sit. now + + + + + + + + + + + + + + + Property in the CoR Discrimination experience – – – – – – – – – – – – – – Knowledge of other CoR regions + + + Knowledge of the CoR language No or poor knowledge (ref.) Intermediate knowledge Good or very good knowledge + + + + + + + + + +

204

The relationship between transnationalism and integration

Sociodemographic and migration- Full model (M11) related variables (M2)

Social Identific. Identific. Social Identific. Identific. integration with CoR with ToR integration with CoR with ToR CoR partner + + + + + + Social contacts in the CoR from CoO + – from 3rd country + + + + + + Indicators below this line were not included in M2 Social contacts in other countries from CoO in CoO from CoO in 3rd country + from 3rd country living abroad Partner lives in other country – – – Knowledge of a 3rd language (at least moderate) Knowledge of 3rd countries Previous migration experience Frequency of transnat. communication via phone and VoIP – via mail and email via social netw. sites Frequency of remittances Received money from other country Property in CoO – – – – Property in 3rd country – Number of trips to CoO Number of trips to 3rd country Voted in last CoO election Consumption of 3rd lang. TV + content Shopping abroad Received e-messages from + + abroad Identification with CoO – – –

Observations 1,329 1,326 1,325 1,184 1,183 1,181 Adjusted R² 0.208 0.116 0.237 0.127 Pseudo R² 0.104 0.121 The used + and – symbols signal whether the respective coefficients are positive or negative while their number indicates the significance level: + + + /– – – p<0.001, + + /– – p<0.01, + /– p<0.05.

With respect to the sociodemographic control variables, it is first of all noteworthy that the effects, in most cases, vary across the different dimensions of integration. Interestingly, group indicators are in the full models irrelevant regarding social integration and mostly irrelevant with respect to identificational integration in the ToR. Hence, it can be concluded from Table 14 that group indicators are broadly relevant regarding CoR identification only, as all Romanian samples differ significantly from the reference group. However, additionally computed regressions (see above) showed that these differences are mostly associated with the origin of the respondents, rather than the general context of immigration in the countries under observation. The two other significant effects, i.e., of Turks in Germany in respect to

205

The relationship between transnationalism and integration identification with the CoR and of Romanians in Italy with respect to ToR identification, occur in the countries which host the biggest migrant communities from these two CoOs in Europe. However, using the data at hand, it is not possible to determine if a connection exists between this fact and the significance of the mentioned indicators. If such a relation existed the mechanisms behind it would differ for both nationalities and would be specific to only one of the three measured aspects of integration in each case. Furthermore, regarding social integration the mentioned effect is not significant anymore in the full model.

Structural integration in the form of a stable household financial situation clearly supports all three analysed aspects of integration. Its effect is slightly more significant in both instances of identificational rather than social integration. While the effect could, theoretically, run in either direction, it can be assumed that in the case of identification, it is relative financial security which impacts the dependent variable and not the other way around. Hence, the better off migrants are, the higher their identification with the country and town they live in, and the higher their probability of having natives as friends and family members.

An influence of cultural integration is most visible in the case of social integration where high knowledge of the CoR language and familiarity with other CoR regions are significant. The respective impact is weaker in the case of ToR identification (in which only good or very good command of the CoR language is significant) and not visible in regard to CoR identification when aspects of transnationalism are controlled for. This means that cultural integration does show a positive impact on the social contact migrants have with the non- migrant population. On the contrary, it does not seem to be directly related to migrants’ self- perception.

The significant effects of perceived discrimination highlight the fact that integration is not only dependent on migrants but also on the public attitudes they encounter in their CoR. Hence, if immigrants do not feel fully accepted in their daily life and have the feeling that they are discriminated against because of their origin they – unsurprisingly – show lower levels of integration in regard to all three indicators observed in the present analysis. The negative effect of discrimination on integration is particularly pronounced in the case of identification with the ToR. This means that experiences of discrimination have the highest impact on identification with the less abstract institutional context and the more immediate social group amongst the two types of identification considered in this analysis. At the same time, occurrences of discrimination have a slightly higher significance for general, and also somewhat emotional, attitudes than on actual behaviour, i.e., social interactions and relations.

206

The relationship between transnationalism and integration

The two different types of social relations which are part of the field of transnational networks (i.e., general contacts and a specific cross-border romantic relationship) have very different effects that are, however, only present in one out of three cases each: Larger networks of CoO acquaintances are positively associated with social integration, while the existence of a partner in another country has a highly significant negative impact on identification with the ToR. First of all, the absence of any negative effects of transnational networks in the social integration dimension underscores how cross-border relationships do not hinder the development of meaningful relationships between migrants and non-migrants in the CoR. The positive indicator can, in fact, be read as a sign that social integration depends more on the general tendency of an individual to socialise with others, than on the national or cultural background those ‘others’ have. This interpretation is strengthened by the highly significant positive effect of the number of friends and family members from third countries that respondents have in the CoR. The negative effect of a partner in another country is confined to the final model of the ToR regression and is not present in the CoR models. As argued above, this might indicate that this aspect is related to the question of whether or not respondents decided to make the ToR the permanent centre of their family life. Apparent is also the absence of an overall negative effect of social integration in the transnational space on indicators of integration into the CoR. Furthermore, negative effects of networks including CoO nationals in the CoR cannot be found. This finding clearly suggests that multiple integration is indeed possible and can be empirically observed.

Only one of the three aspects of transnational communication becomes significant with respect to only one of the three aspects of integration. The fact that the frequency of cross- border communication via phone and VoIP is negatively related to the identificational integration at ToR level is indeed puzzling. In fact, it could be related to the negative effect of having a partner who lives abroad and would strengthen the interpretation that a situation where the respondents’ family life is not centred in the local context, would negatively impact their identification with it.

Economic transnationalism constitutes the only set of practices which has a uniform effect on more than one indicator of integration. Property ownership in the CoO has a strongly negative effect on social integration and identification with the CoR. Furthermore, a negative association exists between the CoR identification and property ownership in third countries. Hence, local identification is the only aspect of CoR integration taken into account in this analysis that is not impacted by property in another country. It is not only striking that

207

The relationship between transnationalism and integration remittances are not significant in all three models, but that the same also holds true for transnational mobility. Hence, having real estate abroad is more influential in regard to integration than frequent visits to the CoO and other countries – which would allow the temporal personal use of such property – and as the act of sending money to friends or family. A possible explanation lies in the fact that real estate is, to some extent, an investment in the future and also a type of immobile capital that can be activated and put to use at a later point. In this sense, the negative effect of having property in the CoR on the tendency to develop social relations to CoR nationals, and on identification with the CoR, could be related to an unspecific hope or plan to return to the CoO in the future.

A significant effect of the identification with the respondents’ CoO can only be found on identificational integration into the CoR. In fact, a highly significant negative relation exists between the two variables. Hence, only in the case of identifications which are situated at the same level, one effectively diminishes the other. However, on the subnational level such a connection cannot be found. This is in accordance with the fact that local contexts are assumed to be more salient for respondents’ daily lives. Ultimately, it is the town and not the country of residence which constitutes the immediate context of migrants’ lives and social interactions. This may make it easier to identify with it and to reconcile this identification with an ethnic or national identification that is directed at the CoO. In this context it is also important, as Schneider et al. (2012, 312-13) point out, that discourses of exclusion very often refer to a national level upon which migrants are presented and styled as ‘the other’. Consequently, the belonging at a local level may have a less exclusive character.

The previous discussion considered the effects of individual indicators of transnationalism separately. In addition to this subchapter, computations of the full models for all three aspects of integration, using the indices discussed in chapter 5 are provided in Appendix 7. On a general level, the models mostly show the same effects. However, the use of indices obscures important details of the results. Considering economic transnationalism as an example, it is in this case not apparent that real estate ownership abroad and not remittances is the main driving force behind the negative relationship between this type of transnationalism and two of the indicators of integration. Two deviations from the previously discussed models are that the index of transnational communication is not significant in any of the regressions and that the cumulative measure of transnational consumption is not only positively related to the identification with the ToR but also with the CoR.

208

The relationship between transnationalism and integration

In conclusion, three key observations result from Table 14 and this discussion: First and most importantly, it can be seen that the influence of transnationalism on social and identificational integration is very limited. However, where effects do exist, they can be of a positive or of a negative nature. This means that there are indeed some aspects of transnationalism which are negatively associated with integration. Nevertheless, these results importantly underline that transnationalism and integration cannot be perceived as mutually exclusive at a general level. Consequently, the findings indicate that multiple integration is indeed possible and can be empirically observed. This implies that assimilation is not the sole possible mode of successful integration in the CoR. Second, in contrast to transnationalism indicators, a comparatively high number of sociodemographic factors are associated with all three aspects of integration. The corresponding effects are relatively stable since they persist even when transnationalism is controlled for. Finally, the presented results strengthen a multidimensional approach to integration. Just as in the case of determinants of transnationalism, it became clear that the same indicators could differ in their effect on the three aspects of integration.

209

7 Conclusion

The main objectives of this study were: first, to provide a detailed cross-national account of the transnationalism of Turkish and Romanian migrants in Denmark, Germany, and Italy; second, to investigate the determinants of their transnational engagement; and, finally, to examine the relationship between transnationalism and integration.

To accomplish the first goal, the analysis took a variety of transnationalism indicators into consideration. In doing so, this study provides a valuable addition to the literature, widening the usual focus applied by research on migrant transnationalism. A particularly important aspect in this regard was the use of a broader geographic lens. While most studies of migrant transnationalism concentrate on activities and social networks that involve individuals in the country of residence (CoR) and country of origin (CoO), this work went beyond this dichotomy. It did so by using data on networks and activities that connect migrants to a variety of countries. This approach was strongly influenced by recent research on transnationalism amongst mostly sedentary populations (e.g., Mau 2010).

The decision to use a broad measurement of transnationalism by including a comparatively high number of indicators took the results of earlier studies into consideration. These studies found that the determinants of transnationalism and also the impact of transnationalism on integration differed depending on the nature of the transnational aspect or activity taken into account (Snel, Engbersen, and Leerkes 2006; Dekker and Siegel 2013; Schunck 2014). To allow for analytical clarity, a multidimensional understanding of transnationalism was used that differentiated between three main dimensions and five subdimensions. Adapting a systematisation proposed by Kuhn (2011), transnational human capital, transnational background and networks, and transnational practices served as the main dimensions. The systematic inclusion of transnational human capital, and transnational background and networks allowed differentiating their effects from those of transnational practices. This was helpful given that the effects of social networks could, for example, otherwise be mistakenly attributed to transnational communication (i.e., if the networks were not controlled for). Furthermore, the inclusion of the two additional dimensions besides transnational practices allows relating the results more directly to existing and future research on transnationalism in sedentary populations. While it is not surprising that migrants possess a certain amount of transnational human capital, and mostly have contacts in other countries, it is nevertheless important to measure those dimensions in order to identify possible differences compared to non-migrant populations. Such differences are possible regarding the existence of

210

Conclusion transnational human capital and background, and their role as determining factors of transnational practices. Within the dimension of transnational practices, the study differentiated between the five subdimensions of transnational communication, economic transnationalism, transnational mobility, political transnationalism, and transnational consumption. This systematisation built on similar ones used in migration research (e.g., Itzigsohn et al. 1999; Portes, Guarnizo, and Landolt 1999; Snel, Engbersen, and Leerkes 2006). While indicators of transnationalism were assigned to different dimensions and subdimensions based on previous research and theoretical reasoning, their relation was empirically confirmed using multiple correspondence analysis (MCA).

The first research question asked to which extent migrants in Denmark, Germany, and Italy engage in transnationalism. In this regard, the study confirmed earlier results, which showed a high degree of variation depending on the specific aspect of transnationalism under observation. An apparent majority phenomenon in all samples is the existence of social contacts (i.e., friends or family members) in other countries. Even in the sample with the lowest rate of social contacts (Turkish migrants in Germany) almost three out of four respondents are part of transnational social networks. This is in line with figures reported by Fauser et al. (2015, 1502), who found that 80 percent of immigrants participating in the German SOEP had contacts abroad. The high rates found in the present study are not surprising given that the sample only included individuals who were born in another country and not naturalised in the CoR. The CoO is the most important location for respondents’ transnational contacts. An essential contribution of this research to the field consists in the finding that the transnational networks of the surveyed migrants are, however, in no way limited to connections between migrants’ CoR and CoO. In fact, with the exception of Turks in Germany, the majority of respondents in all samples reported contacts in third countries. Even amongst Turks in Germany, this was true for nearly half the sample. Similar to the results of Mau and collaborators on the non-migrant population in Germany (Mau and Mewes 2007, 213-15; Mau 2007, 99-101), the analysis revealed that the transnational networks of the surveyed migrants reflect historical patterns to some degree. In the present case, this mostly refers to the fact that networks of Turkish and Romanian migrants focus, besides the CoO, on countries that host major communities of their co-nationals due to migratory movements during the 20th century. This means that many Romanians have friends and family in Spain and Italy, while many Turkish migrants have such contacts in Germany and other Northern European countries.

211

Conclusion

Transnational mobility is one of the activities that migrants of both surveyed nationalities frequently engage in. Overall, between 86 and 93 percent of the respondents in all samples undertook at least one trip to another country during the 24 months prior to the survey. As in the case of transnational networks, the CoO was a focal point for this activity. Comparisons with the literature show that respondents did indeed visit their CoO more often than the participants in surveys conducted in the United States (Guarnizo, Portes, and Haller 2003, 1226) and Australia (O'Flaherty, Skrbis, and Tranter 2007, 828). This is, however, not unexpected given that the costs for travels to the CoO are likely to be lower for migrants within Europe than for most migrants in the United States and Australia. The significance of transnational mobility is also apparent if more frequent trips are taken into account with the majority of participants in three samples (Romanians in Denmark and Germany, and Turks in Italy) taking three or more trips abroad within the 24-month period prior to the survey. In the remaining samples, this held true for a quarter or more of the respondents.

Importantly, the results show that solely focusing on transnational mobility between the CoR and CoO obscures the true extent of this practice, since a majority in all but one sample (Turks in Germany) also visited third countries. In general, the countries closest to the CoR were the most likely to be visited. Countries bearing large communities of either nationality were amongst those visited by large portions of each sample reconfirming the link between geographic aspects of this activity and historical migration patterns. The fact that Turkish and Romanian respondents living in Denmark were more likely to have undertaken three or more trips to third countries as opposed to their CoO in the 24 months prior to the survey stresses the relevance of the settlement context. Denmark is a comparatively small country with an excellent infrastructure and is well connected to its neighbouring countries by land, bridges, and ferries. Consequently, these countries are easy to reach from almost any point in Denmark. At the same time, of the three CoRs, Denmark is furthest away from Turkey and Romania.

Besides their intrinsic value, the results regarding the geographic dimension of transnationalism have considerable methodological implications. These findings reiterate that analyses limited to the CoR-CoO dichotomy risk underestimating the degree of migrants’ transnationalism. Beyond the established topics of migration research, overcoming this limitation is particularly important in light of the growing academic interest in forced migration and the settlement of large numbers of refugees in Europe and worldwide. The nature of these migratory movements makes it probable that individuals (involuntarily)

212

Conclusion participating in them will have family members who were also forced to migrate but may have found refuge in different countries. Consequently, it is very likely that these individuals will form geographically diverse transnational networks which may only include few or no people in their CoO. The geographic scope of transnationalism is also relevant if this research subject is understood in a broader sense to include mostly sedentary populations. In this context, a wider lens is a necessity since the transnationalism of these individuals is by definition not directed at the CoO.

Another practice that most of the surveyed Turkish and Romanian migrants engage in is transnational communication, with nearly all respondents who have contacts in other countries participating in it. On average, respondents are in contact with friends or family in another country at least once a week, most frequently using direct verbal communication (traditional telephone or VoIP). This is in line with earlier assessments, according to which the easy accessibility and low costs of these modern means of communication allow migrants and non-migrants to stay in touch and to remain part of each other’s lives over vast geographical distances (Faist, Fauser, and Reisenauer 2013, 51).

The level of engagement in transnational consumption appears inconsistent. In some samples, very few respondents engage in cross-border shopping, watch foreign-language TV content, or regularly receive electronic messages from abroad. At the same time, some of these aspects are majority phenomena in the other samples. In terms of economic transnationalism, the picture is somewhat clearer, with only a minority of respondents in all samples being involved in it. Concerning remittances, the findings in the literature differ profoundly. While in general terms remittance rates amongst the surveyed migrants are similar to average values reported for other European countries, e.g., by Snel et al. (2006, 292), they are considerably lower than those published for migrants in the United States (Itzigsohn and Giorguli Saucedo 2002, 776; Waldinger 2008, 26; Castañeda, Morales, and Ochoa 2014, 320). While rarely considered in the literature (for exceptions see: Snel, Engbersen, and Leerkes 2006; Carling and Hoelscher 2013), real estate abroad can be seen as an important aspect of economic transnationalism. In this analysis, real estate in the CoO was the only transnational feature negatively related to two aspects of integration, namely, social integration and identificational integration at the CoR level.

Political transnationalism appears to be another phenomenon that only a small minority of respondents engage in. However, the findings in respect to this dimension are limited by the fact that just one indicator (absentee voting) was available.

213

Conclusion

The second major research question concerns the determinants of transnational activities. Again, in line with earlier results, the study showed that they differ depending on the type of transnationalism taken into consideration. Some of the predictors (e.g., educational levels) relate to up to three of the observed categories of transnational practices, but several effects are specific to particular subdimensions. Associations between higher formal education and transnational activities are positive in those cases where they do exist. Moreover, individuals who hold higher educational titles are more likely to communicate frequently with friends and family in other countries, to engage in transnational mobility and in cross-border consumption.

While aspects of structural integration were not in the focus of this analysis, they were included as independent variables. In the models investigating the determinants of transnational activities some of these indicators were significantly related to those practices that are most dependent on material resources. More specifically the results show that individuals who evaluate their household’s financial situation more positively and who are integrated into the labour market, are more likely to engage in economic transnationalism and transnational mobility. Furthermore, real estate ownership in the CoR is positively related to transnational mobility. However, regarding property, it is not only the positive relationship with transnational mobility that is noteworthy but also the absence of a negative association between this indicator and economic transnationalism. This directly relates to the popular claim, made amongst others by Huntington (2004, 272), that foreign investments and remittances realised by migrants decrease their ability and willingness to spend and invest money domestically. The present study does not find any indications for such a mechanism. While the association of economic transnationalism and transnational mobility with aspects of structural integration is most prominent, some aspects are also positively related to different activities. For instance, there is a positive relationship between transnational communication and a stable financial situation of respondents’ households, as well as between property ownership in the CoR and transnational consumption.

These results suggest that economic transnationalism and transnational mobility are resource- dependent activities, in accordance with the typology proposed by Itzigsohn and Giorguli Saucedo (2002). On the contrary, there is no support for the reactive transnationalism hypothesis, whereby migrants “engage in transnational practices as a result of their lack of satisfaction with their life in the country of reception.” (Itzigsohn and Giorguli Saucedo 2002, 772). Besides low structural integration, the authors mention discrimination experiences as a

214

Conclusion possible indicator of this category. However, the data shows no significant effects of discrimination on any of the measured dimensions of transnational practices. In this regard, the findings are in line with Carling and Hoelscher (2013, 954) and Schunck (2014, 224, 248), who did not find any effects of discrimination on the frequency of remittances or visits to the CoO.75 This is in contrast to the recent findings of Snel et al. (2016, 525), who reported indications of reactive transnationalism amongst three migrant groups in Rotterdam.

Previous studies have shown that the origin of migrants influences the degree of their engagement in transnational activities (Snel, Engbersen, and Leerkes 2006; Waldinger 2008; Bilgili 2014a). Consequently, the study took advantage of the cross-national design of the used data to test for group effects. This was done by creating indicators that combined the respondents’ CoO and CoR (i.e., Turks in Denmark, Romanians in Denmark and so on). The descriptive analysis showed that respondents in the different samples indeed participate to varying degrees in the measured transnational activities. However, regression models that investigated the determinants of these activities only revealed a strong influence of group indicators in the dimension of transnational consumption. In the remaining four categories such relations were mainly absent, with one group indicator being significant in the dimension of transnational communication only. These results do not contradict the descriptive analysis. They mean, however, that the apparently existent differences between the samples can neither be explained by contextual effects in the CoR nor by aspects associated with the migrants CoO. Sociodemographic differences, transnational human capital and transnational networks were, in fact, more significant in their impact on four of the five subdimensions of transnational practices than group differences. In the case of transnational consumption, i.e., the one practice in which group differences clearly exist, those effects appear to be mostly related to the CoR.

The analysis confirmed the expected relations between transnational practices and the two additionally included dimensions of transnational human capital, and transnational background and social networks. The expansion of the usual focus on transnational practices present in current migration research is especially necessary if the CoR-CoO dichotomy should be transcended. In this situation, it is helpful to control for knowledge of third languages (i.e., a component of transnational human capital) independently of the general measure of formal education. Aspects of transnational background and social networks are significantly related to all measured categories of transnational practices. Engagement in

75 Schunck did find an effect of discrimination on the amount remitted; however, no equivalent measurement was included in this study.

215

Conclusion political transnationalism (i.e., absentee voting) is, for instance, more likely if the respondents have a life partner in another country, which is likely to be the CoO in most cases. The same holds true for economic transnationalism. The latter also shows a positive association with having a higher number of friends and family members from the CoO living in third countries. This could suggest that engagement in economic cross-border practices is related to the prestige migrants may gain within their social networks by being transnationally active in the economic sphere. On the other hand, it could also be a sign of perceived social pressure resulting from peers within the network who remit money or invest in real estate in the CoO. Based on research that shows that the existence of parents (Carling and Hoelscher 2013, 950) or a partner and children in the CoO (Schunck 2014, 248) can be positively related to remitting behaviour, it was hypothesised that the number of friends and family members in the CoO would have a similar effect on economic transnationalism. While this was not found, a positive impact of having a partner abroad was apparent. First and foremost, this suggests that for transnational economic activities involving the CoO, the type of interpersonal relationships to contacts in this country is more important than their quantity.

Transnational communication shows a positive relationship with having contacts of third country origin in a country other than the CoR. At the same time, this practice is positively associated with knowledge of a third country, as an aspect of transnational human capital. This suggests that the engagement of migrants in transnational communication is the stronger, the more diverse their border-crossing social networks are. This finding also reiterates the value of an approach which does not limit the understanding of transnationalism to phenomena which connect the CoO and CoR or people of either origin. Aspects of transnational human capital and transnational mobility also show a strong relation with each other.

The third and final major research question addressed the possible relationship between transnationalism and integration. From a comparative point of view, the descriptive analysis of social integration and identificational integration in the country and town of residence showed clear differences between all six samples. Overall, the majority of respondents in all samples showed some degree of social integration, as measured by the quantity of CoR friends and family members in their social networks within the CoR. The relative share of respondents for whom this holds true is higher in all Turkish samples compared to the Romanian ones. In contrast, identification with the CoR is, on average, higher amongst the Romanians. However, this value is generally low amongst all respondents. Identification with

216

Conclusion the town of residence (ToR) is substantially higher than identification with the CoR in all samples. This is a significant finding, as migrants’ identificational integration is mostly discussed in national terms. However, the results of this study paint a more complex picture and suggest that limiting the analysis to the CoR level could lead to an underestimation of migrants’ identificational integration. Finally, it has been shown that Romanians identify more strongly with their ToR than Turks, except in Denmark, where both samples show the same level of local identification.

Regarding the association of integration and transnationalism, two key observations can be made. First, very few indicators of transnationalism are significantly related to social and identificational integration (see below). Additionally, some aspects of structural integration also positively impact a limited number of transnational practices, as the analysis of determinants of transnationalism showed. Most indicators of transnationalism do indeed not show any significant relation with social integration or identificational integration at the CoR and ToR level. However, several sociodemographic indicators and features of structural and cultural integration influence these three indicators of integration.

The second observation is that the visible effects of transnationalism on integration tend to differ between integration dimensions. The only effect evident in two of the three aspects of integration, namely with respect to social integration and identification with the CoR, is the negative impact of real estate ownership in the CoO. In the case of identification with the CoR, it is complemented by a corresponding negative effect of real estate ownership in a third country. However, neither indicator is significantly associated with identificational integration at the local level. Remarkable is the fact that no corresponding negative impact of transnational mobility is apparent in the computed models. This means that the negative impact of property abroad is independent of frequent visits to the CoO and personal use of these holdings. Furthermore, sending of remittances, which is another activity that falls into the category of economic transnationalism, has no significant impact on any of the types of integration. These results could mean that the negative effect of property abroad is rather connected to an (imagined) prospective than a specific current use. In this sense, this type of real estate may indicate the existence of a vague hope or plan to remigrate to the CoO at some point in the distant future.

A significant positive association exists between the number of third country friends living in a third country and social integration in the CoR. At the same time, none of the other two indicators of transnational social networks are negatively related to this aspect of integration.

217

Conclusion

This means that, in contrast to the prediction which follows from Esser’s arguments (Esser 2001, 21; Esser 2003, 8-9), multiple integration, i.e., simultaneous integration in transnational social networks and in social groups within the CoR, is indeed possible. The positive impact of a high number of third country friends and family members within the CoR on social integration further strengthens this assessment. Consequently, it would appear that the crucial question regarding social integration is less with whom individuals socialise, but whether they generally tend to socialise at all.

Another aspect of social life, namely having a partner in another country, is negatively related to the respondents’ degree of identification with the town they live in. This means that respondents whose family life is not entirely centred in the CoR tend to identify less with the locality they live in. The effect could be imagined in both directions, on the one hand, migrants who are separated by international borders from their partners might show a lesser degree of identification. While, on the other hand, individuals who have some reason not to feel as part of their town of residence may be less inclined to relocate their loved ones. The data show that identification with the ToR is furthermore positively related with transnational consumption. However, it is unlikely that a causal link exists between both phenomena.

The visible negative effects of transnationalism on integration can of course not be neglected. Nevertheless, most of the transnationalism indicators in this study do not show any effect on integration, and there are also positive associations with other transnational features. Overall, this puts these findings in line with the predominant empirical assessment that integration and transnationalism are not mutually exclusive phenomena (e.g., Guarnizo, Portes, and Haller 2003, 1238; Snel, Engbersen, and Leerkes 2006, 304; Bilgili 2014a, 300; Schunck 2014, 286- 87). This implies that assimilation in Esser’s sense does not constitute the only possibility to achieve successful integration in the CoR.

As mentioned, various aspects that do not fall into the realm of transnationalism exert influence on integration. Two of these effects are consistent across all three types of integration. These are the positive impact of a more stable financial situation and the negative effect of perceived discrimination. The latter underlines that integration is a process that depends on two parties and not solely on the attitude or capabilities of immigrants.

Consistent effects of the group indicators were only apparent for identificational integration in the CoR. The analysis showed that respondents in all Romanian samples were significantly more likely to show a higher degree of identification with the CoR than the comparison group

218

Conclusion

(Turks in Denmark). Additional computations indicated that these effects are more likely to be associated with the differences in nationality than in CoR. It is likely that this higher level of identification is related to Romanians status as EU citizens, which may make it easier for them to relate to the EU member state they live in. Hence, an important difference between the two types of identificational integration consists in the effects of group indicators, which are largely absent in the case of ToR identification. Furthermore, identification with the CoO is negatively related to identification with the CoR but not to identification with the ToR. This indicates that migrants may combine identifications more easily when they refer to different structural levels. In other words, with respect to identificational integration at the national level, multiple integration seems indeed unlikely. However, it should be kept in mind that only migrants who had not been naturalised in their CoR were eligible to participate in the survey. It is possible that this population differs systematically in their identificational integration from migrants who did take on the nationality of the country they live in.

The main findings of this study can be summarised as follows: first, certain aspects of transnationalism are widespread amongst migrants, while others are minority phenomena. Most migrants engage, for instance, in transnational communication, but comparatively few exhibit political transnationalism by casting a vote in CoO elections. Second, transnationalism does not appear to be a phenomenon predominantly to be found amongst migrants in disadvantaged social or material positions. Especially discrimination experience does not show significant effects on the studied practices. This is an important finding because it suggests that transnationalism does not function as a coping mechanism for the surveyed migrants. Third, the analysis emphasises that migrant transnationalism is not limited to the CoR-CoO dichotomy. Consequently, an argument has been made to broaden the geographic scope of transnationalism research to avoid underestimating the phenomenon. Fourth, aspects of transnational human capital, and transnational background and networks exert an influence on transnational activities. Again, regarding future research this calls for the use of a study design that systematically includes these dimensions. Incorporating these important features will allow for a more comprehensive comparison of transnationalism amongst migrant and non-migrant populations in the future. Fifth, few indicators of transnationalism are significantly related to social and identificational integration. These aspects of integration appear to be more strongly influenced by a variety of socio-economic features and structural integration aspects. This suggests that the different dimensions of integration are more strongly influenced by each other than by migrants’ transnationalism. Sixth, transnationalism effects that could be found varied amongst the three integration aspects of social integration,

219

Conclusion identificational integration at the CoR level, and identificational integration at the ToR level. Together with the fact that the determinants of transnationalism also differ between categories, this stresses the need to operationalise both phenomena in a multidimensional manner. Seventh, in respect to the specific effects of transnationalism, the data suggest that multiple social integration is possible. In fact, having friends and family members from the CoO in third countries is positively related to social integration in the CoR. Consequently, the inclusion in transnational networks and socialising with the CoR population do not impede each other. On the contrary, property in the CoO is a feature of economic transnationalism that is negatively related to social integration and identification with the CoR. The overall conclusion resulting from findings five, six, and seven is that transnationalism and integration do not appear to be mutually exclusive phenomena. Eighth, by taking two aspects of identificational integration into focus, the study showed that identification with the CoR and ToR are influenced by different mechanisms and that both types of identification are not congruent with each other. The nature of these differences should be further investigated. The fact that migrants in all samples show a higher degree of identification with their town rather than their country of residence implies that studies investigating migrants’ identification should not solely focus on the national level. Ninth, identificational integration at the CoR level stands in a negative relation to identification with the CoO. Finally, while group differences regarding all aspects of transnationalism and integration are clearly visible, they do not belong to their main determinants in most cases. It follows that differences in the integration and transnationalism of migrants cannot be correctly addressed by using nationality as the main criteria of analysis.

Even though the EUCROSS data allows a systematic review of transnationalism in a cross- national setting, this study has, of course, some shortcomings. First and foremost, the target population was, for methodological reasons, limited to migrants who were not naturalised in their CoR. It is possible that this population differs systematically from individuals who immigrated from the same CoOs but took on the citizenship of the country they live in. This is especially likely with regard to identificational integration. Due to the cross-sectional character of the data, it is, furthermore, not possible to arrive at a comprehensive assessment of the development of transnationalism and its relation with integration over time. The indicator of the time respondents spent in their CoR (i.e., duration of stay) can only serve as a limited proxy in this regard. At the same time, it has to be stressed that the goal of this study was a cross-national comparison. The collection of detailed longitudinal data on migrant transnationalism in various countries on specific populations would undoubtedly be desirable.

220

Conclusion

Unfortunately, the author is not aware of any such data set. Finally, the study could not give a detailed review of political transnationalism. The assessment of political transnationalism was hampered by the fact that there was only a single indicator included in the EUCROSS survey that could be used to operationalise the phenomenon.

Further research is necessary regarding the geographic dimension of migrant transnationalism. Migration research should not limit itself to the CoR-CoO dichotomy but investigate the degree of migrants’ presence in more extensive networks and their engagement in activities that connect them to a variety of countries. While additional quantitative data is needed in this regard, a mixed-method approach would be preferable. Such data would allow a deeper understanding of the underlying mechanisms contributing to transnational phenomena.

Furthermore, transnationalism research should take non-migrant populations into view to a higher degree. Comparative data would be of interest not just regarding transnationalism, but also its association with integration. After all, the degree of integration cannot only change for those who have migrated to a country but also for those who were born there. Just as transnationalism is a phenomenon that potentially concerns the whole population, integration is.

221

Appendix

222

Appendix 1: Classification of educational titles76 Classification of Romanian educational Turkish educational titles Danish educational titles German educational titles Italian educational titles educational titles titles

1 Lower secondary Primary education not Primary education not Primary education not Primary education not Primary education not education or less completed; completed; completed; completed; completed; Şcoală primară; İlkokul mezunu; Folkeskole (1. til 9. Volks-/ Licenza elementare; Gimnaziu Ortaokul mezunu klasse) Hauptschulabschluss bzw. Licenza media/ Polytechnische avviamento professionale Oberschule mit Abschluss 8. oder 9. Klasse

2 In-between lower and Şcoală profesională ori de Meslek eğitimi veya Kort erhvervsuddannelse; Mittlere Reife/Realschul- n.a. higher secondary meserii çıraklık Faglig uddannelse abschluss bzw. education Polytechnische Oberschule mit Abschluss 10. Klasse; Fachhochschulreife (Abschluss mit Fachoberschule)

3 Higher secondary Liceu; Lise mezunu Gymnasiel uddannelse, Abitur bzw. Erweiterte Diploma di scuola media education (university Şcoală post-liceală studenter eksamen Oberschule mit Abschluss superiore entrance (inclusiv colegiu) 12. Klasse requirement) (Hochschulreife)

4 Tertiary education Studii Lisans; Kort videregående Fachhochschul- oder Diploma universitario/ superioare/facultate; Master; uddannelse; Hochschulabschluss; laurea di primo livello; Studii post-universitare Doktora Mellemlang videregående Promotion (Dr.; PhD); Laurea vecchio uddannelse; ordinamento/ Lang videregående laurea di secondo livello uddannelse. (specialistica, magistrale); Diploma post-laurea (dottorato, specializzazione)

76 Further details regarding the measurement of educational titles in the EUCROSS study are provided in Pötzschke (2015, 17-30).

Appendix

Appendix 2: Official population statistics on Turkish and Romanian migrants in Denmark, Germany, and Italy

N % of foreign Rank % of total population population Turkish citizens in … Denmark 28,755 7.67 1 0.51 Germany 1,575,717 21.84 1 1.93 Italy 17,711 0.40 34 0.03 Romanian citizens in … Denmark 12,374 3.30 8 0.22 Germany 205,026 2.84 6 0.25 Italy 933,354 21.27 1 1.56 Sources: Denmark (Statistics Denmark 2016), Germany (Statistisches Bundesamt 2013: 26, 40), Italy (ISTAT 2016a; ISTAT 2016b).

Appendix 3: Bivariate relation between migration motives and year of settlement (logistic regression, odds- ratio)77 Work Education Quality of life Family N

Year of settlement of TU in DK 0.94 *** 1.04 0.96 1.06*** 248 TU in DE 0.94 *** 0.99 1.09* 1.06*** 250 TU in IT 0.98 1.22*** 0.95 0.96 247

RO in DK 1.07** 1.18*** 0.96* 0.94** 244 RO in DE 1.04* 1.05 0.94** 1.00 246 RO in IT 0.98 1.01 0.99 1.04 246 *** p<0.001, ** p<0.01, * p<0.05. Additional control: survey mode (all models). EUCROSS, 2016.

77 The table summarises the results of a total of 24 regression models. Each model used one of the migrant motives as dependent variable and the year of settlement as an independent variable. Only the survey mode was used as additional control variable.

224

Appendix 4: Transnational networks – most often mentioned countries (percent) TU in DK TU in DE TU in IT RO in DK RO in DE RO in IT

% of the % of the % of the % of the % of the % of the sample sample sample sample sample sample Turkey 88 Turkey 70 Turkey 86 Romania 84 Romania 70 Romania 71 Germany 62 Netherlands 24 Germany 65 Italy 45 Italy 28 Germany 33 Sweden 33 France 23 France 34 Spain 36 Spain 26 Spain 25 Netherlands 23 Austria 13 Switzerland 15 Germany 35 USA 23 France 18 France 22 Belgium 10 Netherlands 14 USA 22 Canada 17 USA 18 Norway 17 UK 6 USA 14 France 14 France 13 UK 12 Austria 14 Switzerland 6 Austria 13 UK 14 Austria 6 Canada 10 Belgium 12 UK 10 Canada 12 UK 6 Austria 8 Switzerland 8 Belgium 7 Austria 8 Denmark 5 Sweden 7 Spain 5 Greece 6 EUCROSS, 2016. N = 1,502 Appendix 5: Transnational mobility – most often mentioned countries (percent) TU in DK TU in DE TU in IT RO in DK RO in DE RO in IT

% of the % of the % of the % of the % of the % of the sample sample sample sample sample sample Turkey 88 Turkey 83 Turkey 76 Romania 84 Romania 77 Romania 87 Germany 35 France 13 Germany 35 Germany 41 Italy 28 Austria 20 Sweden 20 Netherlands 11 France 27 Italy 17 Austria 22 France 18 Netherlands 8 Austria 8 Switzerland 14 France 17 Spain 22 Germany 14 Norway 6 Austria 12 Austria 16 France 19 Spain 11 Austria 6 Greece 8 Sweden 14 Hungary 11 Hungary 10 France 6 Netherlands 8 Spain 14 Netherlands 8 UK 6 Belgium 5 Spain 7 Hungary 12 Switzerland 8 Belgium 5 Norway 8 Belgium 5 Netherlands 8 UK 5 UK 7 Czech 6 Republic Belgium 5

Greece 5

EUCROSS, 2016. N = 1,502

Appendix

Appendix 6: Independent variables used in regression models Variable Categories/ values Modifications for regression analyses

Group 1= Turkish migrants in Denmark; Recoded into dummy variables 2 = Turkish migrants in Germany; 3 = Turkish migrants in Italy; 4 = Romanian migrants in Denmark; 5 = Romanian migrants in Germany; 6 = Romanian migrants in Italy Gender (female) 0 = no; 1 = yes Civil status 1 = married/ registered civil union; Recoded into dummy variables 2 = single, never been married; 3 = other Age at migration In years Duration of stay In years Education 1 = lower secondary education or less; Recoded into dummy variables 2 = in-between lower and higher secondary education; 3 = higher secondary education (university entrance requirement); 4 = tertiary education Received highest education in 0 = no; 1 = yes CoR Labour market participation 1 = working (in part or full-time); Recoded into dummy variables 2 = non-working; 3 = retired; 4 = unemployed; 5 = other Household financial situation 1 = we were living very comfortably on Reverse coded at age 14 the money we had; 2 = we were living 1 = we found it very difficult … comfortably on the money we had; 3 = 5 = we were living very comfortably we made ends meet; 4 = we found it on the money we had difficult; 5 = we found it very difficult Household financial situation 1= we are living very comfortably on Reverse coded now the money we have; 2 = we are living 1 = we find it very difficult … comfortably on the money we have; 5 = we are living very comfortably on 3 = we make ends meet; 4 = we find it the money we have difficult; 5 = we find it very difficult Property in the CoR 0 = no; 1 = yes Discrimination experience 0 = no; 1 = yes Knowledge of other CoR 1 = one; 2 = two or more; 3 = no Recoded regions 0 = no; 1 = one; 2 = two or more Knowledge of the CoR 1= Speaks CoR language almost as Recoded language well as native language; 2 = quite well; 0 = no or poor knowledge; 3 = just so-so; 4 = poorly; 5 = not at all; 1 = intermediate knowledge; 2 = good or very good knowledge CoR partner 0 = no; 1 = yes Social contacts in the CoR - 1 = a lot; 2 = a few; 3 = none Reverse coded from CoO 0 = none; 1 = a few; 2 = a lot Social contacts in the CoR - 1 = a lot; 2 = a few; 3 = none Reverse coded from 3rd country 0 = none; 1 = a few; 2 = a lot Social contacts in other 1 = a lot; 2 = a few; 3 = none Reverse coded countries - from CoO in CoO 0 = none; 1 = a few; 2 = a lot Social contacts in other 1 = a lot; 2 = a few; 3 = none Reverse coded countries - from CoO in 3rd 0 = none; 1 = a few; 2 = a lot country

226

Appendix

Variable Categories/ values Modifications for regression analyses

Social contacts in other 1 = a lot; 2 = a few; 3 = none Reverse coded countries - from 3rd country 0 = none; 1 = a few; 2 = a lot living abroad Partner lives in other country 1 = yes; 2 = no; Reverse coded 0 = no; 1 = yes Knowledge of a 3rd language 1= Speaks add. language almost as well Recoded into a dummy variable (at least moderate) as native language; 2 = quite well; 0 = never studied an additional 3 = just so-so; 4 = poorly; 5 = not at all; language, not at all, poorly; 1 = just 6 = never studied an additional so-so, quite well, speaks add. language language almost as well as native language Knowledge of 3rd countries 1 = one; 2 = two or more; 3 = no Recoded 0 = no; 1 = one; 2 = two or more Previous migration 0 = no; 1 = yes experience Frequency of transnational 0 = never; 1 = less often than once a communication - via phone month; 2 = at least once a month; and VoIP 3 = at least once a week; 4 = every day Frequency of transnational 0 = never; 1 = less often than once a communication - via mail and month; 2 = at least once a month; email 3 = at least once a week; 4 = every day Frequency of transnational 0 = never; 1 = less often than once a communication - via social month; 2 = at least once a month; networking sites 3 = at least once a week; 4 = every day Frequency of remittances 0 = never; 1 = less than once a year; 2 = at least once a year; 3 = at least once a month Received money from other 0 = no; 1 = yes country Property in CoO 0 = no; 1 = yes Property in 3rd country 0 = no; 1 = yes Number of trips to CoO 0 = none; 1 = one; 2 = two; 3 = 3-5; 4 = Recoded 6-10; 5 = more than 10 0 = none; 1 = one; 2 = two; 3 = three or more Number of trips to 3rd 0 = none; 1 = one; 2 = two; 3 = 3-5; 4 = Recoded country 6-10; 5 = more than 10 0 = none; 1 = one; 2 = two; 3 = three or more Voted in last CoO election 0 = no; 1 = yes Consumption of 3rd language 0 = never; 1 = less often than once a Recoded into a dummy variable TV content month; 2 = at least once a month; 3 = at 0 = no; 1 = yes least once a week; 4 = every day Shopping abroad 0 = no; 1 = yes Received e-messages from 0 – 100 % Recoded into a dummy variable abroad 0 = no messages from abroad received; 1 = messages from abroad received Identification with CoO 1 = strongly disagree … 5 = strongly agree

227

Appendix

Appendix 7: Regression of social integration and identificational integration using transnationalism indices (coefficients) Social integration Identificational Identificational integration at integration at CoR level ToR level Groups TU in DK (ref.) TU in DE 0.205 -0.373** 0.047 TU in IT -0.697 0.450 0.021 RO in DK -0.209 0.568*** 0.180 RO in DE -0.229 0.966*** 0.157 RO in IT 0.150 0.831*** 0.497** Gender (female = 1) 0.196 0.052 -0.020 Married/ registered civil union (ref.) Single, never been married 0.015 -0.241* -0.095 Other -0.380 0.138 -0.140 Age at migration 0.006 0.002 0.011 Duration of stay 0.016 0.026*** 0.028*** Education Lower secondary or less (ref.) In-between low. and higher second. 0.351 -0.089 -0.364* Higher secondary 0.001 0.073 -0.116 Tertiary 0.179 0.332** -0.041 Received highest education in CoR -0.116 0.245* 0.194 Labour market participation Working (ref.) Non-Working 0.065 -0.019 0.029 Retired 0.087 -0.082 -0.092 Unemployed -0.156 -0.000 -0.094 Other 0.198 -0.102 -0.042 Household financial sit. at age 14 -0.049 -0.105** -0.087 Household financial sit. now 0.210** 0.253*** 0.179** Property in the CoR 0.020 -0.098 -0.005 Discrimination experience -0.272* -0.248*** -0.454*** Knowledge of other CoR regions 0.188** 0.050 0.023 Knowledge of the CoR language No or poor knowledge (ref.) Intermediate knowledge 0.174 0.057 0.110 Good or very good knowledge 0.654** 0.266 0.458* CoR partner 0.894*** 0.120 0.090 Social contacts in the CoR from CoO 0.168 -0.083 -0.023 from 3rd country 0.774*** 0.050 0.081 Social contacts in other countries from CoO in CoO 0.071 -0.028 -0.002 from CoO in 3rd country 0.283** 0.033 0.054 from 3rd country living abroad 0.005 -0.030 -0.085 Partner lives in other country -0.188 -0.191 -0.865*** Knowledge of a 3rd language 0.209 -0.106 -0.075 (at least moderate) Knowledge of 3rd countries 0.018 -0.074 0.032 Previous migration experience -0.185 -0.095 -0.050 Transnational communication (index)78 0.023 0.003 -0.001

78 The base category of the transnational communication index includes, in this case, all respondents who did not report any contacts abroad. This deviation from previously used coding (chapter 5) is necessary because otherwise the regression model would be limited to individuals with such transnational contacts, disregarding all others. While this limitation was logical when communication was used a dependent variable, it is not helpful when it serves as one of several predictors of integration. The effect of the absence of transnational contacts is controlled for through the use of the indicators of transnational background and networks. If the original transnational communication index is used, and the regressions consequently disregard all respondents who do

228

Appendix

Social integration Identificational Identificational integration at integration at CoR level ToR level Economic transnationalism (index) -0.098* -0.074** -0.053 Transnational mobility (index) 0.022 -0.012 0.007 Political transnationalism 0.083 -0.012 -0.146 Transnational consumption (index) 0.147 0.100* 0.200*** Received money from other country -0.123 0.082 0.004 Property in 3rd country -0.357 -0.775* -0.576 Identification with CoO 0.085 -0.153*** 0.037

Observations 1,184 1,183 1,181 Pseudo R-squared 0.116 Adjusted R-squared 0.233 0.122 *** p<0.001, ** p<0.01, * p<0.05. Additional controls (all models): survey mode; information on CoR language knowledge missing by design; did not live in CoR or was not eligible to vote at last CoO election. EUCROSS, 2016

not have any friends and family members abroad, the coefficients of the index are still not significant in any of the three models (additional computations not included in the table).

229

Bibliography

Agnew, John. 1994. "The Territorial Trap. The Geographical Assumptions of International Relations Theory." Review of International Political Economy 1 (1): 53-80.

Aksoy, Asu and Kevin Robins. "Banal Transnationalism. The Difference that Television Makes." Transnational Communities Programme - Working Paper Series WPTC-02-08, Transnational Communities Programme, Oxford, 2002, accessed April 9, 2017, http://www.transcomm.ox.ac.uk/working%20papers/WPTC-02-08%20Robins.pdf.

Alaminos, Antonio and Oscar Santacreu. 2009. "Living Across Cultures in a Transnational Europe." In Pioneers of European Integration. Citizenship and Mobility in the EU, edited by Ettore Recchi and Adrian Favell, 98-119. Cheltenham: Elgar.

Alba, Richard and Nancy Foner. 2016. "Integration's Challenges and Opportunities in the Wealthy West." Journal of Ethnic and Migration Studies 42 (1): 3-22. doi:10.1080/1369183X.2015.1083770.

———. 2015a. "Mixed Unions and Immigrant-Group Integration in North America and Western Europe." The Annals of the American Academy of Political and Social Science 662 (1): 38-56. doi:10.1177/0002716215594611.

———. 2015b. Strangers no more. Princeton: Princeton Univ. Press.

Alba, Richard and Victor Nee. 2003. Remaking the American Mainstream. Assimilation and Contemporary Immigration. Cambridge, Mass.: Harvard Univ. Press.

Aleinikoff, T. Alexander and Douglas Klusmeyer. 2001. "Plural Nationality. Facing the Future in a Migratory World." In Citizenship Today. Global Perspectives and Practices, edited by T. Alexander Aleinikoff and Douglas Klusmeyer, 63-88. Washington, DC: Brookings Inst. Press.

Amelina, Anna, Thomas Faist, Nina Glick Schiller, and Devrimsel D. Nergiz. 2012. "Methodological Predicaments of Cross-Border Studies." In Beyond Methodological Nationalism. Research. Methodologies for Cross-Border Studies, edited by Anna Amelina, Devrimsel D. Nergiz, Thomas Faist and Nina Glick Schiller, 1-19. New York, NY: Routledge.

Anderson, Benedict. 2006. Imagined Communities. Reflections on the Origin and Spread of Nationalism. London: Verso.

Andrén, Daniela and Monica Roman. 2016. "Should I Stay or should I Go? Romanian Migrants during Transition and Enlargements." In Labor Migration, EU Enlargement, and the Great Recession, edited by Martin Kahanec and Klaus F. Zimmermann, 247-269. Berlin: Springer. doi:10.1007/978-3-662-45320-9_11.

Anghel, Remus Gabriel. 2013. Romanians in Western Europe. Migration, Status Dilemmas, and Transnational Connections. Lanham: Lexington Books.

230

Bibliography

Ataman, Ferda and Ghelli, Fabio. "Türkische Präsidentschaftswahl. Ein kompliziertes Verfahren mit kurzen Fristen." Mediendienst Integration, last modified June 24, 2014, accessed September 09, 2017, http://mediendienst-integration.de/artikel/tuerkische- praesidentschaftswahl-deutschland.html.

Bade, Klaus J. 2003. Migration in European History. Malden, Mass.: Blackwell.

Bade, Klaus J. and Michael Bommes. 2004. "Einleitung. Integrationspotentiale in modernen europäischen Wohlfahrtsstaaten — Der Fall Deutschland." In Migrationsreport 2004 Fakten, Analysen, Perspektiven, edited by Klaus J. Bade, Michael Bommes and Rainer Münz, 11-42. Frankfurt/Main: Campus.

Ban, Cornel. 2012. "Economic Transnationalism and its Ambiguities. The Case of Romanian Migration to Italy." International Migration 50 (6): 129-149. doi:10.1111/j.1468- 2435.2009.00556.x.

Bauböck, Rainer. 2006. "Citizenship and Migration. Concepts and Controversies." In Migration and Citizenship. Legal Status, Rights and Political Participation, edited by Rainer Bauböck, 15-31. Amsterdam: Amsterdam Univ. Press.

———. 2005. "Expansive Citizenship. Voting Beyond Territory and Membership." PS: Political Science and Politics 38 (4): 683-687.

———. 2003. "Towards a Political Theory of Migrant Transnationalism." International Migration Review 37 (3): 700-723.

Berry, John W. 1997. "Immigration, Acculturation, and Adaptation." Applied Psychology 46 (1): 5-34. doi:10.1111/j.1464-0597.1997.tb01087.x.

Bilgili, Özge. 2014a. "Migrants’ Multi-Sited Social Lives. Interactions between Sociocultural Integration and Homeland Engagement." Comparative Migration Studies 2 (3): 283-304. doi:10.5117/CMS2014.3.BILG.

———. 2014b. Simultaneity in Transnational Migration Research. Links between Migrants’ Host and Home Country Orientation. Maastricht: Boekenplan.

Blasius, Jörg. 2010. "Korrespondenzanalyse." In Handbuch der sozialwissenschaftlichen Datenanalyse, edited by Christof Wolf and Henning Best, 367-389. Wiesbaden: VS Verlag für Sozialwissenschaften.

Boccagni, Paolo. 2012. "Rethinking Transnational Studies. Transnational Ties and the Transnationalism of Everyday Life." European Journal of Social Theory 15 (1): 117-132. doi:10.1177/1368431011423600.

Bochove, Marianne van, Katja Rusinovic, and Godfried Engbersen. 2010. "The Multiplicity of Citizenship. Transnational and Local Practices and Identifications of Middle-Class Migrants." Global Networks 10 (3): 344-364. doi:10.1111/j.1471-0374.2010.00292.x.

Bommes, Michael. 2012. "Transnationalism or Assimilation?" In Immigration and Social Systems. Collected Essays of Michael Bommes, edited by Christina Boswell and Gianni D'Amato, 107-124. Amsterdam: Amsterdam Univ. Press.

231

Bibliography

Bouras, Nadia. 2013. "Shifting Perspectives on Transnationalism. Analysing Dutch Political Discourse on Moroccan Migrants' Transnational Ties, 1960–2010." Ethnic and Racial Studies 36 (7): 1219-1231. doi:10.1080/01419870.2013.785580.

Braun, Michael and Walter Müller. 2012. "National and Transnational Identities of Intra- European Migrants." In Crossing Borders, Shifting Boundaries. National and Transnational Identities in Europe and Beyond, edited by Franz Höllinger and Markus Hadler, 263-287. Frankfurt/Main: Campus.

Braun, Michael and Ettore Recchi. 2009. "Free-Moving Western Europeans. An Empirically Based Portrait." In Migration and Mobility in Europe. Trends, Patterns and Control, edited by Heinz Fassman, Max Haller and David Lane, 85-101. Cheltenham: Elgar.

Brubaker, Rogers. 2001. "The Return of Assimilation? Changing Perspectives on Immigration and its Sequels in France, Germany, and the United States." Ethnic and Racial Studies 24 (4): 531-548. doi:10.1080/01419870120049770.

Bundesministerium für Justiz und für Verbraucherschutz (BMJV). "Staatsangehörigkeitsgesetz (StAG)." Bundesministerium für Justiz und für Verbraucher- schutz, 2015, http://www.gesetze-im-internet.de/rustag/BJNR005830913.html.

Carling, Jø and Kristian Hoelscher. 2013. "The Capacity and Desire to Remit. Comparing Local and Transnational Influences." Journal of Ethnic and Migration Studies 39 (6): 939-958. doi:10.1080/1369183X.2013.765657.

Castañeda, Ernesto, Maria Cristina Morales, and Olga Ochoa. 2014. "Transnational Behavior in Comparative Perspective. The Relationship between Immigrant Integration and Transnationalism in New York, El Paso, and Paris." Comparative Migration Studies 2 (3): 305-333. doi:10.5117/CMS2014.3.CAST.

Center for Migration and Development (CMD). "The Comparative Immigrant Entrepreneurship Project (CIEP)." Princeton University, 2017, accessed March 12, 2016, https://cmd.princeton.edu/publications/data-archives/ciep.

Central Intelligence Agency (CIA). "Suffrage." The World Factbook, Central Intelligence Agency, 2017, accessed February 18, 2017, https://www.cia.gov/library/publications/the- world-factbook/fields/2123.html.

Crul, Maurice and Jens Schneider. 2010. "Comparative Integration Context Theory. Participation and Belonging in New Diverse European Cities." Ethnic and Racial Studies 33 (7): 1249-1268. doi:10.1080/01419871003624068.

Crul, Maurice and Hans Vermeulen. 2003. "The Second Generation in Europe." International Migration Review 37 (4): 965-986. doi:10.1111/j.1747-7379.2003.tb00166.x.

Dahinden, Janine. 2009. "Are we all Transnationals Now? Network Transnationalism and Transnational Subjectivity. The Differing Impacts of Globalization on the Inhabitants of a Small Swiss City." Ethnic and Racial Studies 32 (8): 1365-1386.

232

Bibliography

Dalakoglou, Dimitris. 2010. "Migrating-Remitting-‘Building’-Dwelling. House-Making as ‘Proxy’ Presence in Postsocialist Albania." Journal of the Royal Anthropological Institute 16 (4): 761-777. doi:10.1111/j.1467-9655.2010.01652.x.

De Haas, Hein. 2005. "International Migration, Remittances and Development. Myths and Facts." Third World Quarterly 26 (8): 1269-1284.

Dekker, Bram and Melissa Siegel. "Transnationalism and Integration. Complements or Substitutes?" UNU‐MERIT Working Paper Series 71, UNU‐MERIT, 2013, accessed January 15, 2016, http://digitalarchive.maastrichtuniversity.nl/fedora/get/guid:bd7c3e01- 35cf-4327-9310-785251472dc4/ASSET1.

Del Boca, Daniela and Alessandra Venturini. 2005. "Italian Migration." In European Migration. What do we Know?, edited by Klaus F. Zimmermann, 303-335. Oxford: Oxford Univ. Press.

Díez Medrano, Juan and Paula Gutiérrez. 2001. "Nested Identities. National and European Identity in Spain." Journal of Ethnic and Racial Studies 24 (5): 753-778. doi:10.1080/01419870120063963.

Ehrkamp, Patricia. 2005. "Placing Identities. Transnational Practices and Local Attachments of Turkish Immigrants in Germany." Journal of Ethnic and Migration Studies 31 (2): 345-364. doi:10.1080/1369183042000339963.

Elger, Katrin and Gezer, Özlem. "For Erdogan, or Against Him. Turks in Germany Get the Vote." Spiegel Online, last modified August 6, 2014, accessed September 7, 2015, http://www.spiegel.de/international/germany/turks-in-germany-allowed-to-vote-for-first- time-a-984624.html.

Engbersen, Godfried, Marek Okólski, Richard Black, and Cristina Pantîru. 2010. "Working out a Way from East to West. EU Enlargement and Labour Migration from Central and Eastern Europe." In A Continent Moving West? EU Enlargement and Labour Migration from Central and Eastern Europe, edited by Richard Black, Godfried Engbersen, Marek Okólski and Cristina Pantîru, 7-22. Amsterdam: Amsterdam Univ. Press.

Erdal, Marta Bivand. 2012. "‘A Place to Stay in Pakistan’. Why Migrants Build Houses in their Country of Origin." Population, Space and Place 18 (5): 629-641. doi:10.1002/psp.1694.

Erdal, Marta Bivand and Ceri Oeppen. 2013. "Migrant Balancing Acts. Understanding the Interactions between Integration and Transnationalism." Journal of Ethnic and Migration Studies 39 (6): 867-884. doi:10.1080/1369183X.2013.765647.

Escobar, Cristina. 2010. "La contribution des organisations de migrants Latino-Américains des États-Unis au développement de leurs pays d'origine." Critique Internationale 46 (1): 103-124.

———. 2007. "Extraterritorial Political Rights and Dual Citizenship in Latin America." Latin American Research Review 42 (3): 43-75.

233

Bibliography

Esser, Hartmut. 2003. "Ist das Konzept der Assimilation überholt?" Geographische Revue 5 (2): 5-22.

———. "Integration und ethnische Schichtung." Arbeitspapiere 40, Mannheimer Zentrum für Europäische Sozialforschung, Mannheim, 2001, accessed September 25, 2015, http://www.mzes.uni-mannheim.de/publications/wp/wp-40.pdf.

———. 2000. Soziologie. Spezielle Grundlagen. Die Konstruktion der Gesellschaft, Hartmut Esser. Vol. 2. Frankfurt/Main: Campus-Verlag.

———. 1980. Aspekte der Wanderungssoziologie. Assimilation und Integration von Wanderern, ethnischen Gruppen und Minderheiten. Eine handlungstheoretische Analyse. Darmstadt: Luchterhand.

European Commission. "Report from the Commission to the Council on the Functioning of the Transitional Arrangements on Free Movement of Workers from Bulgaria and Romania." COM(2011) 729, 2011, accessed October 24, 2013, http://ec.europa.eu/ social/BlobServlet?docId=7204&langId=en.

EUROSTAT. "Migrants in Europe. A Statistical Portrait of the First and Second Generation." Statistical books, Publications Office of the European Union, Luxembourg, 2011, accessed October 07, 2016, http://ec.europa.eu/eurostat/documents/3217494/ 5727749/KS-31-10-539-EN.PDF.

Faist, Thomas. 2013. "The Mobility Turn. A New Paradigm for the Social Sciences?" Ethnic and Racial Studies 36 (11): 1637-1646. doi:10.1080/01419870.2013.812229.

———. 2012. "Toward a Transnational Methodology. Methods to Address Methodological Nationalism, Essentialism, and Positionality." Revue Européenne des Migrations Internationales 28 (1): 51-70.

———. 2010. "Diaspora and Transnationalism. What Kind of Dance Partners?" In Diaspora and Transnationalism. Concepts, Theories and Methods, edited by Rainer Bauböck and Thomas Faist, 9-34. Amsterdam: Amsterdam Univ. Press.

———. 2001. "Integration(en) von Immigranten aus Polen in Deutschland. Assimilation, Pluralismus und transstaatliche Integration." In Polen in Bremen eine unsichtbare Minderheit?, edited by Zdzisław Krasnode̜ bski and Nele Krampen, 27-55. Bremen: Universität Bremen.

———. 2000. The Volume and Dynamics of International Migration and Transnational Social Spaces. Oxford: Clarendon.

Faist, Thomas, Margit Fauser, and Eveline Reisenauer. 2013. Transnational Migration. Cambridge, Malden: Polity Press.

Faist, Thomas and Christian Ulbricht. "Von Integration zu Teilhabe? Anmerkungen zum Verhältnis von Vergemeinschaftung und Vergesellschaftung." COMCAD Working Papers 130, Center on Migration, Citizenship and Development, Bielefeld, 2014, accessed February 5, 2016, http://www.uni-bielefeld.de/tdrc/ag_comcad/downloads/ WP_130.pdf.

234

Bibliography

Fauser, Margit. 2017. "Mixed Methods and Multisited Migration Research. Innovations from a Transnational Perspective." Journal of Mixed Methods Research. doi: 10.1177/1558689817702752.

Fauser, Margit, Elisabeth Liebau, Sven Voigtländer, Hidayet Tuncer, Thomas Faist, and Oliver Razum. 2015. "Measuring Transnationality of Immigrants in Germany. Prevalence and Relationship with Social Inequalities." Ethnic and Racial Studies 38 (9): 1497-1519. doi:10.1080/01419870.2015.1005639.

Fawcett, James T. 1989. "Networks, Linkages, and Migration Systems." The International Migration Review 23 (3): 671-680. doi:10.2307/2546434.

Foner, Nancy. 2007. "Engagements Across National Borders, Then and Now." Fordham Law Review 75 (5): 2483-2492.

Gans, Herbert J. 1992. "Second‐Generation Decline. Scenarios for the Economic and Ethnic Futures of the post‐1965 American Immigrants." Ethnic and Racial Studies 15 (2): 173- 192. doi:10.1080/01419870.1992.9993740.

Gerhards, Jürgen and Silke Hans. 2013. "Transnational Human Capital, Education, and Social Inequality. Analyses of International Student Exchange." Zeitschrift Für Soziologie 42 (2): 99-117. doi:10.1515/zfsoz-2013-0203.

Gerhards, Jürgen, Silke Hans, and Sören Carlson. 2014. "Transnationales Humankapital. Einleitende Bemerkungen zu Kontextbedingungen, Erwerb und Verwertbarkeit von transnationalen Kompetenzen." In Globalisierung, Bildung Und Grenzüberschreitende Mobilität, edited by Jürgen Gerhards, Silke Hans and Sören Carlson, 7-19. Wiesbaden: Springer Fachmedien. doi:10.1007/978-3-658-02439-0_1.

Glasze, Georg and Andreas Pott. 2014. "Räume der Migration und der Migrationsforschung." In Räumliche Auswirkungen der internationalen Migration, edited by Paul Gans, 47-62. Hannover: Verlag der ARL - Akademie für Raumforschung und Landesplanung.

Glick Schiller, Nina. 2005. "Transborder Citizenship. An Outcome of Legal Pluralism within Transnational Social Fields." In Mobile People, Mobile Law. Expanding Legal Relations in a Contracting World, edited by Franz von Benda-Beckmann, Keebet von Benda- Beckmann and Anne M. O. Griffiths, 27-49. Aldershot, Hants: Ashgate.

Glick Schiller, Nina, Linda Basch, and Christina Blanc-Szaton. 1992a. "Towards a Definition of Transnationalism." Annals of the New York Academy of Sciences 645: ix-xiv. doi:10.1111/j.1749-6632.1992.tb33482.x.

———. 1992b. "Transnationalism. A New Analytic Framework for Understanding Migration." Annals of the New York Academy of Sciences 645: 1-24. doi:10.1111/j.1749- 6632.1992.tb33484.x.

Glick Schiller, Nina, Linda Basch, and Christina Szanton Blanc. 1999. "From Immigrant to Transmigrant. Theorizing Transnational Migration." In Migration and Transnational Social Spaces, edited by Ludger Pries, 73-105. Aldershot, Hants: Ashgate.

235

Bibliography

Goldring, Luin. 1999. "The Power of Status in Transnational Social Fields." In Migration and Transnational Social Spaces, edited by Ludger Pries, 162-186. Aldershot, Hants: Ashgate.

Gordon, Milton Myron. 1964. Assimilation in American Life. The Role of Race, Religion, and National Origins. New York: Oxford Univ. Press.

Gosh, Bimal. 2006. Migrants’ Remittances and Development. Myths, Rhetoric and Realities. Geneva: International Organization for Migration.

Greenacre, Michael. 1994. "Correspondence Analysis and its Interpretation." In Correspondence Analysis in the Social Sciences. Recent Developments and Applications, edited by Michael Greenacre and Jörg Blasius, 3-22. London: Acad. Press.

Groenewold, George and Laurence Lessard-Phillips. 2012. "Research Methodology." In The European Second Generation Compared. Does the Integration Context Matter?, edited by Maurice Crul, Jens Schneider and Frans Lelie, 39-56. Amsterdam: Amsterdam Univ. Press.

Guarnizo, Luis E., Alejandro Portes, and William Haller. 2003. "Assimilation and Transnationalism. Determinants of Transnational Political Action among Contemporary Migrants." The American Journal of Sociology 108 (6): 1211-1248.

Guarnizo, Luis E. and Michael Peter Smith. 1998. "The Locations of Transnationalism." In Transnationalism from Below, edited by Michael Peter Smith and Luis E. Guarnizo, 3- 31. New Brunswick: Transaction Publishers.

Habermas, Jürgen. "Leadership and Leitkultur." Opinion, The New York Times, last modified October 28, 2010, accessed May 26, 2016, http://nyti.ms/1D3vPhF.

Herbert, Ulrich. 2003. Geschichte der Ausländerpolitik in Deutschland. Saisonarbeiter, Zwangsarbeiter, Gastarbeiter, Flüchtlinge. Bonn: C. H. Beck.

Hiebert, Daniel and David Ley. 2006. "Characteristics of Immigrant Transnationalism in Vancouver." In Transnational Identities and Practices in Canada, edited by Vic Satzewich and Lloyd L. Wong, 71-90. Vancouver: UBC Press.

Humpert, Andreas and Klaus Schneiderheinze. 2000. "Stichprobenziehung für telefonische Zuwandererumfragen. Einsatzmöglichkeiten der Namensforschung." ZUMA Nachrichten 24 (47): 36-63.

Huntington, Samuel P. 2004. Who are we? Die Krise der amerikanischen Identität. Hamburg; Wien: Europaverlag.

Icduygu, Ahmet and Ibrahim Sirkeci. 1998. "Changing Dynamics of the Migratory Regime between Turkey and Arab Countries." Turkish Journal of Population Studies 20: 3-13.

Icduygu, Ahmet, Ibrahim Sirkeci, and Gü Muradoglu. 2001. "Socio-Economic Development and International Migration. A Turkish Study." International Migration 39 (4): 39-61. doi:10.1111/1468-2435.00162.

236

Bibliography

International Institute for Democracy and Electoral Assistance (IDEA). "Compulsory Voting." International Institute for Democracy and Electoral Assistance (IDEA), 2017, accessed September 30, 2017, https://www.idea.int/data-tools/data/voter-turnout/ compulsory-voting.

International Organization for Migration (IOM). "Global Migration Trends Factsheet 2015." International Organization for Migration, Global Migration Data Analysis Centre, Berlin, 2016, accessed March 8, 2017, http://publications.iom.int/books/global-migration-trends- factsheet-2015.

Istituto nazionale di statistica (ISTAT). "Demographic Balance for the Year 2012 and Resident Population on 31st December 2012." Demo, Istituto nazionale di statistica, 2016a, accessed April 1, 2016, http://demo.istat.it/bil2012/index_e.html.

———. "Foreign Citizens. Resident Population by Sex and Citizenship on 31st December 2012." Demo, Istituto nazionale di statistica, 2016b, accessed April 1, 2016, http://demo.istat.it/str2012/index_e.html.

Itzigsohn, José. 1995. "Migrant Remittances. Labor Markets, and Household Strategies. A Comparative Analysis of Low-Income Household Strategies in the Caribbean Basin." Social Forces 74 (2): 633-655.

Itzigsohn, José, Carlos Dore Cabral, Esther Hernandez Medina, and Obed Vazquez. 1999. "Mapping Dominican Transnationalism. Narrow and Broad Transnational Practices." Ethnic and Racial Studies 22 (2): 316-339.

Itzigsohn, José and Silvia Giorguli Saucedo. 2002. "Immigrant Incorporation and Sociocultural Transnationalism." International Migration Review 36 (3): 766-798.

Itzigsohn, José and Daniela Villacrés. 2008. "Migrant Political Transnationalism and the Practise of Democracy. Dominican External Voting Rights and Salvadoran Home Town Associations." Ethnic and Racial Studies 31 (4): 664-686.

Jones-Correa, Michael. 2001. "Under Two Flags. Dual Nationality in Latin America and its Consequences for Naturalization in the United States." International Migration Review 35 (4): 997-1029.

Karakaşoğlu, Yasemin. 2008. "Türkische Arbeitswanderer in West-, Mittel- und Nordeuropa Seit der Mitte der 1950er Jahre." In Enzyklopädie Migration in Europa. Vom 17. Jahrhundert bis zur Gegenwart, edited by Klaus J. Bade, Pieter C. Emmer, Leo Lucass and Jochen Oltmer, 1054-1061. Paderborn: Schöningh.

Kastoryano, Riva. 2002. Negotiating Identities States and Immigrants in France and Germany. Princeton Studies in Cultural Sociology. Princeton, NJ: Princeton Univ. Press.

Kivisto, Peter and Thomas Faist. 2009. Beyond a Border. The Causes and Consequences of Contemporary Immigration. Los Angeles: Pine Forge Press.

Koehn, Peter H. and James N. Rosenau. 2002. "Transnational Competence in an Emergent Epoch." International Studies Perspectives 3 (2): 105-127. doi:10.1111/1528- 3577.00084.

237

Bibliography

Kritz, Mary M. and Hania Zlotnik. 1992. "Global Interactions. Migration Systems, Processes, and Policies." In International Migration Systems a Global Approach, edited by Mary M. Kritz, Lin Lean Lim and Hania Zlotnik, 1-16. Oxford: Clarendon Press.

Kuhn, Theresa. 2015. Experiencing European Integration Transnational Lives and European Identity. Oxford: Oxford Univ. Press.

———. 2011. "Individual Transnationalism, Globalisation and Euroscepticism. An Empirical Test of Deutsch's Transactionalist Theory." European Journal of Political Research 50 (6): 811-837. doi:10.1111/j.1475-6765.2011.01987.x.

Lacroix, Thomas. 2013. "Collective Remittances and Integration. North African and North Indian Comparative Perspectives." Journal of Ethnic and Migration Studies 39 (6): 1019- 1035. doi:10.1080/1369183X.2013.765667.

Landolt, Patricia. 2001. "Salvadoran Economic Transnationalism. Embedded Strategies for Household Maintenance, Immigrant Incorporation, and Entrepreneurial Expansion." Global Networks 1 (3): 217-242. doi:10.1111/1471-0374.00014.

Lawler, Edward J. 1992. "Affective Attachments to Nested Groups. A Choice-Process Theory." American Sociological Review 57 (3): 327-339.

Levitt, Peggy. 2002. "Keeping Feet in Both Worlds. Transnational Practices and Immigrant Incorporation in the United States." In Toward Assimilation and Citizenship Immigrants in Liberal Nation-States, edited by Christian Joppke and Ewa Morawska, 177-194. Basingstoke: Palgrave Macmillan.

———. 2001. "Transnational Migration. Taking Stock and Future Directions." Global Networks. A Journal of Transnational Affairs 1 (3): 195-216.

Levitt, Peggy and Nina Glick Schiller. 2004. "Conceptualizing Simultaneity. A Transnational Social Field Perspective on Society." International Migration Review 38 (3): 1002-1039. doi:10.1111/j.1747-7379.2004.tb00227.x.

Ley, David. 2013. "Does Transnationalism Trump Immigrant Integration? Evidence from Canada's Links with East Asia." Journal of Ethnic and Migration Studies 39 (6): 921- 938. doi:10.1080/1369183X.2013.765654.

Lockwood, David. 1964. "Social Integration and System Integration." In Explorations in Social Change, edited by George Karl Zollschan and Walter Hirsch, 244-257. London: Routledge & Kegan Paul.

Lucassen, Leo. 2006. "Is Transnationalism Compatible with Assimilation? Examples from Western Europe since 1850." IMIS Beiträge 29: 15-36.

———. 1997. "The Gulf between Long Term and Short Term Approaches in Immigration Studies. A Reassessment of the Chicago School's Assimilation Concept." IMIS Beiträge 5: 5-23.

238

Bibliography

Marcelli, Enrico A. and B. Lindsay Lowell. 2005. "Transnational Twist. Pecuniary Remittances and the Socioeconomic Integration of Authorized and Unauthorized Mexican Immigrants in Los Angeles County." International Migration Review 39 (1): 69-102. doi:10.1111/j.1747-7379.2005.tb00256.x.

Martins, Herminio. 1974. "Time and Theory in Sociology." In Approaches to Sociology. an Introduction to Major Trends in British Sociology, edited by John Rex, 246-294. London: Routledge & Paul.

Mau, Steffen. 2010. Social Transnationalism. Lifeworlds Beyond the Nation-State. London and New York: Routledge.

———. 2007. Transnationale Vergesellschaftung. Die Entgrenzung sozialer Lebenswelten. Frankfurt/Main: Campus-Verl.

Mau, Steffen and Jan Mewes. 2007. "Transnationale soziale Beziehungen. Eine Kartographie der deutschen Bevölkerung." Soziale Welt 58 (2): 203-222. doi:10.5771/0038-6073- 2007-2-203.

Mau, Steffen, Jan Mewes, and Ann Zimmermann. 2008. "Cosmopolitan Attitudes through Transnational Social Practices?" Global Networks 8 (1): 1-24. doi:10.1111/j.1471- 0374.2008.00183.x.

Mazzolari, Francesca. "Determinants of Naturalization. The Role of Dual Citizenship Laws." CCIS Working Papers 117, The Center for Comparative Immigration Studies. University of California, San Diego, 2005, accessed June 19, 2010, http://ccis.ucsd.edu/wp- content/uploads/2012/07/wrkg117.pdf.

Menjivar, Cecilia, Julie DaVanzo, Lisa Greenwell, and R. Burciaga Valdez. 1998. "Remittance Behavior among Salvadoran and Filipino Immigrants in Los Angeles." The International Migration Review 32 (1): 97-126. doi:10.2307/2547562.

Migazin. "So haben Türken in Deutschland gewählt." Migazin, last modified June 8, 2015, accessed June 9, 2015, http://www.migazin.de/2015/06/08/wahlergebnisse-tuerkei-so- tuerken-deutschland/.

Morawska, Ewa. 2003. "Disciplinary Agendas and Analytic Strategies of Research on Immigrant Transnationalism. Challenges of Interdisciplinary Knowledge." International Migration Review 37 (3): 611-640.

———. 2002. "Immigrant Transnationalism and Assimilation. A Variety of Combinations and the Analytic Strategy it Suggests." In Toward Assimilation and Citizenship Immigrants in Liberal Nation-States, edited by Christian Joppke and Ewa Morawska, 133-176. Basingstoke: Palgrave Macmillan.

Mügge, Liza. 2016. "Transnationalism as a Research Paradigm and its Relevance for Integration." In Integration Processes and Policies in Europe. Contexts, Levels and Actors, edited by Blanca Garcés-Mascareñas and Rinus Penninx, 109-125. Cham: Springer.

239

Bibliography

Müller-Mahn, Detlef. 2000. "Ein ägyptisches Dorf in Paris. Eine empirische Studie zur Süd- Nord-Migration am Beispiel ägyptischer ›Sans-papiers‹ in Frankreich." IMIS Beiträge, (15), 79-110.

Nannestad, Peter. 2004. "Immigration as a Challenge to the Danish Welfare State?" European Journal of Political Economy 20: 755-767. doi:10.1016/j.ejpoleco.2004.03.003.

O'Flaherty, Martin, Zlatko Skrbis, and Bruce Tranter. 2007. "Home Visits. Transnationalism among Australian Migrants." Ethnic and Racial Studies 30 (5): 817-844. doi:10.1080/01419870701491820.

Oğuz, Gönül. 2012. EU Enlargement and Turkish Labour Migration. Tokyo: United Nations Univ. Press.

Okyay, Aslı. 2015. "Diaspora-Making as a State-Led Project. Turkey's Expansive Diaspora Strategy and its Implications for Emigrant and Kin Populations." PhD diss., European University Institute. doi:10.2870/151027.

———. "External Voting to be Implemented for the First Time in Turkey’s Forthcoming Presidential Elections." European Union Democracy Observatory on Citizenship, European University Institute, 2014, accessed September 7, 2015, http://eudo- citizenship.eu/news/citizenship-news/1080-external-voting-to-be-implemented-for-the- first-time-in-turkey-s-forthcoming-presidential-elections.

Østergaard-Nielsen, Eva. 2003. "The Politics of Migrants’ Transnational Political Practices." International Migration Review 37 (3): 760-786.

Park, Robert E. 1921. "The Assimilation of Races." In Introduction to the Science of Sociology, edited by Robert E. Park and Ernest W. Burgess, 756-762. Chicago, Ill.: The University of Chicago press.

Park, Robert E. and Ernest W. Burgess. 1921. "Assimilation. Introduction." In Introduction to the Science of Sociology, edited by Robert E. Park and Ernest W. Burgess, 734-740. Chicago, Ill.: The University of Chicago press.

Pauli, Julia. 2008. "A House of One's Own. Gender, Migration, and Residence in Rural Mexico." American Ethnologist 35 (1): 171-187. doi:10.1111/j.1548-1425.2008.00012.x.

Pedersen, Peder J. 2005. "Migration in the Scandinavian Welfare State. The Recent Danish Experience." In European Migration. What do we Know?, edited by Klaus F. Zimmermann, 59-87. Oxford: Oxford Univ. Press.

Pedersen, Peder J. and Nina Smith. "International Migration and Migration Policy in Denmark." CLS Working Paper 01-05, Centre for Labour Market and Social Research, Aarhus, 2001, accessed April 10, 2016. doi: 10.2139/ssrn.298580.

Penninx, Rinus and Blanca Garcés-Mascareñas. 2016. "Integration Processes and Policies in Europe. Contexts, Levels and Actors." In Integration Processes and Policies in Europe. Contexts, Levels and Actors, edited by Blanca Garcés-Mascareñas and Rinus Penninx, 11-29. Cham: Springer.

240

Bibliography

Popescu, Bogdan. 2012. "Out-of-Country Voting. Participation in Elections of Romanian Diaspora." Romanian Review of Political Sciences and International Relations 9 (1): 94- 110.

Portes, Alejandro. 2009. "Migration and Development. Reconciling Opposite Views." Ethnic and Racial Studies 32 (1): 5-22.

———. 2007. "Migration, Development, and Segmented Assimilation. A Conceptual Review of the Evidence." The Annals of the American Academy of Political and Social Science 610 (1): 73-97. doi:10.1177/0002716206296779.

———. 2003. "Conclusion. Theoretical Convergencies and Empirical Evidence in the Study of Immigrant Transnationalism." International Migration Review 37 (3): 874-892. doi:10.1111/j.1747-7379.2003.tb00161.x.

———. 1999. "Conclusion. Towards a New World. The Origins and Effects of Transnational Activities." Ethnic and Racial Studies 22 (2): 463-477.

Portes, Alejandro, Luis E. Guarnizo, and William J. Haller. 2002. "Transnational Entrepreneurs. An Alternative Form of Immigrant Economic Adaptation." American Sociological Review 67 (2): 278-298.

Portes, Alejandro, Luis E. Guarnizo, and Patricia Landolt. 1999. "The Study of Transnationalism. Pitfalls and Promise of an Emergent Research Field." Ethnic and Racial Studies 22 (2): 217-237.

Portes, Alejandro and Min Zhou. 1993. "The New Second Generation. Segmented Assimilation and its Variants." The Annals of the American Academy of Political and Social Science 530: 74-96.

Potot, Swanie. 2010. "Transitioning Strategies of Economic Survival. Romanian Migration during the Transition Process." In A Continent Moving West? EU Enlargement and Labour Migration from Central and Eastern Europe, edited by Richard Black, Godfried Engbersen, Marek Okólski and Cristina Pantîru, 249-269. Amsterdam: Amsterdam Univ. Press.

Pötzschke, Steffen. 2016. "Doppelte Staatsangehörigkeit und Auslandswahlrecht in Kolumbien." In Migration und Demokratie, edited by Stefan Rother, 161-183. Wiesbaden: Springer VS.

———. "Technical Report of the Quantitative EUCROSS Survey." EUCROSS Working Paper 9, 2015, accessed November 5, 2016, http://www.eucross.eu/cms/index.php? option=com_docman&task=doc_download&gid=91&Itemid=157.

Pries, Ludger. 2010. Transnationalisierung. Theorie und Empirie grenzüberschreitender Vergesellschaftung. Wiesbaden: VS Verlag für Sozialwissenschaften.

———. 2008. Die Transnationalisierung der sozialen Welt. Sozialräume jenseits von Nationalgesellschaften. Frankfurt/Main: Suhrkamp.

241

Bibliography

———. 1996. "Transnationale soziale Räume. Theoretisch-empirische Skizze am Beispiel der Arbeitswanderungen Mexico-USA". Zeitschrift für Soziologie, 25(6), 456-471.

Pries, Ludger and Martin Seeliger. 2012. "Transnational Social Spaces. Between Methodological Nationalism and Cosmo-Globalism." In Beyond Methodological Nationalism Research. Methodologies for Cross-Border Studies, edited by Anna Amelina, Devrimsel D. Nergiz, Thomas Faist and Nina Glick Schiller, 219-238. New York, NY: Routledge.

Recchi, Ettore, Michael Braun, Adrian Favell, Juan Díez Medrano, Mike Savage, Laurie Hanquinet, and Dumitru Sandu (EUCROSS). 2016. The Europeanisation of Everyday Life. Cross-Border Practices and Transnational Identities among EU and Third-Country Citizens (EUCROSS). Data file Version 1.0.0 ed. Vol. ZA5911. Cologne: GESIS Data Archive. doi:10.4232/1.12381.

Rodríguez-García, Dan, Dan Rodríguez-García, Miranda J. Lubbers, Miguel Solana, and Verónica de Miguel-Luken. 2015. "Contesting the Nexus between Intermarriage and Integration. Findings from a Multi-Dimensional Study in Spain." The Annals of the American Academy of Political and Social Science 662 (1): 223-245. doi:10.1177/0002716215598136.

Safar, Hayssam, Alain Modot, Silvia Angrisani, Yves Gambier, Carlo Eugeni, Héloïse Fontanel, Najwa Hamaoui, and Xavier Verstrepen. "Study on the Use of Subtitling. The Potential of Subtitling to Encourage Foreign Language Learning and Improve the Mastery of Foreign Languages ." European Commission, 2011, accessed April 7, 2017, http://eacea.ec.europa.eu/llp/studies/documents/study_on_the_use_of_subtitling/rapport_ final-en.pdf.

Sandu, Dumitru. 2005. "Dynamics of Romanian Emigration After 1989. From a Macro- to a Micro-Level Approach." International Journal of Sociology 35 (3): 36-56.

Schans, Djamila. 2009. "Transnational Family Ties of Immigrants in the Netherlands." Ethnic and Racial Studies 32 (7): 1164-1182. doi:10.1080/01419870902763852.

Schneider, Jens and Maurice Crul. 2010. "New Insights into Assimilation and Integration Theory. Introduction to the Special Issue." Ethnic and Racial Studies 33 (7): 1143-1148. doi:10.1080/01419871003777809.

Schneider, Jens, Tineke Fokkema, Raquel Matias, Snežana Stojčić, Dušan Ugrina, and Constanza Vera-Larrucea. 2012. "Identities. Urban Belonging and Intercultural Relations." In The European Second Generation Compared. Does the Integration Context Matter?, edited by Maurice Crul, Jens Schneider and Frans Lelie, 285-340. Amsterdam: Amsterdam Univ. Press.

Schunck, Reinhard. 2014. Transnational Activities and Immigrant Integration in Germany. Concurrent or Competitive Processes? Cham: Springer.

Schütz, Alfred. 1945. "On Multiple Realities." Philosophy and Phenomenological Research 5 (4): 533-576.

242

Bibliography

Serrano Carrasco, Ángela Lucía. 2003. "Colombia. La posibilidad de una ciudadanía sin fronteras." In Votar en la Distancia. La extensión de los derechos políticos a migrantes, Experiencias comparadas, edited by Leticia Calderón Chelius, 115-144. México, D.F.: Instituto Mora.

Silberman, Roxane, Richard Alba, and Irène Fournier. 2007. "Segmented Assimilation in France? Discrimination in the Labour Market Against the Second Generation." Ethnic and Racial Studies 30 (1): 1-27. doi:10.1080/01419870601006488.

Smith, Lothar and Valentina Mazzucato. 2009. "Constructing Homes, Building Relationships. Migrant Investments in Houses." Tijdschrift voor economische en sociale geografie 100 (5): 662-673. doi:10.1111/j.1467-9663.2009.00571.x.

Smith, Robert C. 2003. "Diasporic Memberships in Historical Perspective. Comparative Insights from the Mexican, Italian and Polish Cases." International Migration Review 37 (3): 724-759.

Snel, Erik, Margrietha 't Hart, and Marianne van Bochove. 2016. "Reactive Transnationalism. Homeland Involvement in the Face of Discrimination." Global Networks 16 (4): 511- 530. doi:10.1111/glob.12125.

Snel, Erik, Godfried Engbersen, and Arjen Leerkes. 2006. "Transnational Involvement and Social Integration." Global Networks 6 (3): 285-308. doi:10.1111/j.1471- 0374.2006.00145.x.

Song, Miri. 2009. "Is Intermarriage a Good Indicator of Integration?" Journal of Ethnic and Migration Studies 35 (2): 331-348. doi:10.1080/13691830802586476.

Spierings, Nils. 2014. "Electoral Participation and Intergenerational Transmission among Turkish Migrants in Western Europe." Acta Politica online first. doi:10.1057/ap.2014.26.

Stan, Sabina and Roland Erne. 2014. "Explaining Romanian Labor Migration. From Development Gaps to Development Trajectories." Labor History 55 (1): 21-46. doi:10.1080/0023656X.2013.843841.

Statistics Denmark. "Population at the First Day of the Quarter by Region, Citizenship and Time." StatBank Denmark, Statistics Denmark, 2016, accessed March 23, 2016, http://www.statbank.dk/folk1.

Statistisches Bundesamt. 2013. Statistisches Jahrbuch 2013. Wiesbaden: Statistisches Bundesamt.

Stepick, Alex and Carol Dutton Stepick. 2010. "The Complexities and Confusions of Segmented Assimilation." Ethnic and Racial Studies 33 (7): 1149-1167. doi:10.1080/01419871003599518.

Tager, Michael. 2006. "Expatriates and Elections." Diaspora. A Journal of Transnational Studies 15 (1): 35-60.

243

Bibliography

Tubergen, Frank van, Ineke Maas, and Henk Flap. 2004. "The Economic Incorporation of Immigrants in 18 Western Societies. Origin, Destination, and Community Effects." American Sociological Review 69 (5): 704-727. doi:10.1177/000312240406900505.

Üçok, Mine. "Transnational Consumption Practices for Social Mobility. A Study of Turkish Immigrants in Denmark ." AMID Working Paper Series 58, The Academy for Migration Studies in Denmark/Aalborg University, Aalborg, 2006, accessed November 1, 2016, http://www.amid.dk/pub/.

Urry, John. 2007. Mobilities. Cambridge, Malden: Polity Press.

Vermeulen, Hans. 2010. "Segmented Assimilation and Cross-National Comparative Research on the Integration of Immigrants and their Children." Ethnic and Racial Studies 33 (7): 1214-1230. doi:10.1080/01419871003615306.

Vertovec, Steven. 2009. Transnationalism. London: Routledge.

———. 2007. "Super-Diversity and its Implications." Ethnic and Racial Studies 30 (6): 1024- 1054. doi:10.1080/01419870701599465.

———. 2004. "Cheap Calls. The Social Glue of Migrant Transnationalism." Global Networks 4 (2): 219-224. doi:10.1111/j.1471-0374.2004.00088.x.

Waldinger, Roger. 2010. "Unacceptable Realities. Public Opinion and the Challenge of Immigration in a Franco-American Comparison." In Migration in a Globalised World, edited by Cédric Audebert and Mohamed Kamel Doraï, 41-61. Amsterdam: Amsterdam Univ. Press.

———. 2008. "Between “Here” and “There”. Immigrant Cross-Border Activities and Loyalties." International Migration Review 42 (1): 3-29. doi:10.1111/j.1747- 7379.2007.00112.x.

Wimmer, Andreas and Nina Glick Schiller. 2003. "Methodological Nationalism, the Social Sciences, and the Study of Migration. An Essay in Historical Epistemology." The International Migration Review 37 (3): 576-610.

———. 2002. "Methodological Nationalism and Beyond. Nation–state Building, Migration and the Social Sciences." Global Networks 2 (4): 301-334. doi:10.1111/1471- 0374.00043.

Worbs, Susanne, Eva Bund, Martin Kohls, and Christian Babka von Gostomski. "(Spät-) Aussiedler in Deutschland. Eine Analyse aktueller Daten und Forschungsergebnisse." Forschungsbericht 20, Bundesamt für Migration und Flüchtlinge, Nürnberg, 2013, accessed March 25, 2016, http://www.bamf.de/SharedDocs/Anlagen/DE/ Publikationen/Forschungsberichte/fb20-spaetaussiedler.pdf;jsessionid=291B528DB05E0 8D4F57F89552D420FA3.1_cid368?__blob=publicationFile.

World Bank. "Personal Remittances, Paid (Current US$)." World Bank Open Data, World Bank, 2015a, accessed September 3, 2015, http://data.worldbank.org/indicator/ BM.TRF.PWKR.CD.DT.

244

Bibliography

———. "Personal Remittances, Received (% of GDP)." World Bank Open Data, World Bank, 2015b, accessed September 3, 2015, http://data.worldbank.org/indicator/ BX.TRF.PWKR.DT.GD.ZS.

245